MDCU 3000747 — Final Project Review

1 · Assignment Brief

3000747 Healthcare Innovation and Entrepreneurship — Group Project (40% weight)

Students transition from "Clinical Science Students" to "Healthcare Startup Co-Founders". Working in teams of 2 (suggested), they identify a real-world clinical / patient-care / digital-health / medtech / therapeutic problem, develop a solution, validate it with real users, and build a Minimum Viable Product (MVP). They pitch the venture to a panel demonstrating both clinical/technical feasibility and business viability.

Deliverables checklist (from the brief, page 3)

#	Deliverable	Specification
1	Pitch Deck (PDF)	10–12 slides excluding cover; animation permitted; ≤5 backup pages
2	Validation Evidence	≥2 potential users interviewed / demoed per student
3	Demo Video	5–7 minutes showing problem, solution, prototype in action
4	MVP	Prototype or equivalent brought to showcase
5	Final Report	≤1000 words excluding references and infographics
6	Additional Comment Response Document	Separate file explaining post-interim revisions + locations in report

Grading rubric (40% total — verbatim from the brief)

Criterion	Weight	Details
Problem-Solution Fit	10%	Is the problem real? Does the solution actually address the problem based on customer insights?
MVP	5%	Quality of design; feasibility and quality of the prototype
Validation Evidence	5%	Did they conduct user testing / experiment with the user using a prototype? Do they plan to improve features using user feedback?
Business Viability	10%	Evidence of market needs and a viable Business Model
Presentation & Pitch	10%	Clarity, storytelling, and persuasion skills

↓ Download the full assignment brief (PDF)

2 · Committee Interim Feedback

After the interim presentation, the committee delivered written feedback for each project. Five projects were covered: Manopause Chatbot, EV Optima AI, MindCheck, CellGuard, and NutriTrack. The committee rated each "Excellent / Good / Fair" and produced action items grouped by Critical / Major / Minor severity.

Project	Overall	Key Action Items (per committee)
Project 1 Manopause Chatbot	Good	Value & Journey: Add People-Process-Technology framework and Human-in-the-Loop element Scope: Expand beyond menopause — add e-commerce, telehealth, hospital integration Personalization: Clarify what data is retained and how it's secured Business model: Explain revenue capture if patient goes direct to hospital Missing details: Pricing strategy + go-to-market References: APA citations linked to main body
Project 2 EV Optima AI	Excellent	Data acquisition: Outline how/where to source training data beyond own datasets Competitor analysis: Spec-by-spec comparison with quantified time/cost benefits Economic impact: Calculate dollar savings with vs without product Commercialization pathway: Map impact across R&D → Preclinical → Clinical → Regulatory Missing details: Pricing strategy + go-to-market References: APA citations linked to main body
Project 3 MindCheck	Fair	Business model: Clarify revenue capture if patient goes direct to hospital Value & Journey: Add People-Process-Technology + Human-in-the-Loop Scope: Opportunities beyond chatbot — telehealth, hospital integration, vision Competitor analysis: Deeper spec-by-spec comparison Business fundamentals: Pricing + B2B paying customer (universities?) + go-to-market (hospital integration focus) Visuals: Slides too messy — review Visual Aids lecture AI content: Verify AI-generated content before submission References: APA citations linked to main body
Project 4 CellGuard AI	Good	Pain points: Define specific human errors (oxygen drops, heater drying) + calculate ROI Competitor analysis: Spec-by-spec comparison with benefits Future state: Clearly define what "predictive" means for your technology Market positioning: Address concern this looks like internal R&D or equipment-company feature; define unique startup value References: APA citations linked to main body
Project 5 NutriTrack AI	Excellent	Workflow integration: Show exactly how the app fits into daily clinical workflow Caregiver-centric design: Assume low health literacy (Thai caregivers); keep intuitive Competitor analysis: Spec-by-spec comparison with error/time reduction benefits Market sizing: Concrete TAM/SAM/SOM numbers; focus on big cities first Strategic partnerships: Collaborate with dietary product / pharma / nutritionists Missing details: Pricing strategy + go-to-market References: APA citations linked to main body

Chayanee's OonJai Breast Partners (a 6th project surfaced in the final submissions) is not covered in Feedback.docx. The adaptation check for that project is omitted per the grading methodology.

↓ Download the original Committee Feedback (DOCX)

3 · Student Submissions

All files uploaded to MyCourseVille by the 7 May 2026 deadline are listed below, grouped by student. Click any file to download. Each card lists explicit "Missing / Non-compliant" items and tags presentation status.

Cohort status note: 5 of 6 projects have presented in class. OonJai Breast Partners (Chayanee Sae-lim) has not yet presented — the rubric scoring on the Presentation & Pitch criterion remains provisional until the in-class pitch takes place.

6878305730

Xin Cai Solo

Project 1 · Menopause Chatbot (MenoChat)

Demo video · 5:49 ✅

Missing / Non-compliant All 6 brief-listed deliverables submitted and format-compliant. Video 5:49 within 5–7 min spec. Deck 11 content slides + 4 backup within 10–12 / ≤5 spec.

Presented in class

6878303430

Chadaporn Attakitbancha Solo

Project 2 · EV Optima AI

Pitch Deck (revised)pdf
Validation Board Resultsxlsx
Final Reportdocx
Additional Comment Response Documentdocx
↗ Live MVP · ev-boost-insight.lovable.appweb

MCV text-box content Brief — video link only

Submitted on 02-May-2026 @ 10:46:15 — text-box was used only to share the demo video link (the substantive Comment Response is in Additional_Comment_Response_Document.docx).

Demo Video: drive.google.com/file/d/11Lr2dao2ZsBPwEiFSl23LGGcjhI9JSfU/view

Demo video · 4:01 ⚠ under 5 min

Missing / Non-compliant

Demo video under-duration: 4:01 vs 5–7 min spec (content quality strong though).
Pitch deck slightly over slide cap: 13 content slides vs 10–12 limit.

All 6 deliverables present; no missing files.

Presented in class

6678313530

Xiaoyan Li Solo

Project 3 · MindCheck (Mental health triage)

Note: Final Report and separate Comment Response Document were not provided as separate files — content embedded in the MCV text-box (shown in full below).

MCV text-box — full content De-facto Comment Response

Submitted on 06-May-2026 @ 04:37:43 from 49.229.162.30 · The entire comment-response and revision narrative was placed in the rich-text-box rather than as a separate file.

Demo link: drive.google.com/file/d/1B83QqoLQ7Gs5SiQGpN8sEl1qtEGSMGwV/view

"Thank you to all the professors for your valuable suggestions; I have incorporated these improvements into the PPT accordingly."

1. Business Model

The business model has been revised in the PPT based on the feedback to clearly explain how value is captured. The platform generates revenue through referral fees, SaaS subscriptions, and B2B partnerships. Clinics are charged ฿100–300 per confirmed visit when users are referred through the platform, and a monthly SaaS fee of ฿1,000–฿5,000 for access to pre-screened patients and operational tools. Universities pay subscription fees for mental health insights and prevention programs. As addressed in the feedback, even if users go directly to hospitals, the platform still captures value through SaaS by improving patient quality, triage efficiency, and conversion rates.

2. Value and User Journey (People – Process – Technology)

The PPT has been updated to incorporate the People–Process–Technology framework as suggested. The key stakeholders (People) include students, licensed therapists, universities, and clinics. The Process now clearly reflects the care pathway: screening using PHQ-9/GAD-7, AI-based risk stratification, and appropriate recommendations, followed by professional intervention for moderate to high-risk cases. Importantly, the human-in-the-loop element has been emphasized, where licensed professionals are responsible for final diagnosis and treatment. The Technology layer includes AI emotional journaling, risk assessment algorithms, referral systems, and data dashboards.

3. Beyond Chatbot (Long-term Vision)

Based on the feedback, the PPT now explicitly highlights opportunities beyond the chatbot. The solution is positioned as a scalable digital mental healthcare platform, including telehealth services, integration with hospitals and clinics, and a mental health marketplace. Additional B2B services for universities and institutions have also been included. The long-term vision has been clarified as building a full-stack ecosystem that connects users, providers, and institutions.

4. Competitor Analysis

The competitor analysis has been strengthened in the PPT with a more detailed, feature-by-feature comparison. It highlights that the platform provides structured clinical screening (PHQ-9/GAD-7), risk-based pathways, decision support, and clinical triage with referral capabilities. In contrast, competitors such as OOCA focus on direct therapy without screening or decision guidance. The revised analysis clearly emphasizes the unique value proposition: guiding users from awareness to appropriate care, rather than only providing therapy.

5. Missing Business Fundamentals

The PPT has been updated to address the missing business fundamentals identified in the feedback. The pricing strategy is now clearly defined, including referral fees, SaaS subscriptions, university plans, and optional premium features for students. The target paying customers are clarified, with clinics as the primary revenue source and universities as secondary customers, while students are positioned as users rather than main payers. The go-to-market strategy has also been refined to include clinic partnerships, university pilot programs, campus promotion, and scaling through hospital integration, as recommended.

Demo video · 4:27 ⚠ under 5 min

Missing / Non-compliant

❌ Final Report — no separate file provided.
❌ Comment Response Document — no separate file; content sits inside the MCV text-box (5 numbered points).
⚠ Validation evidence — deck describes a future "PILOT PLAN" rather than completed user testing.
⚠ Pitch deck format — PPTX submitted, brief asks for PDF.
⚠ Demo video under-duration — 4:27 vs 5–7 min spec.

Presented in class

6871004530

Zinab Omer Solo

Project 4 · CellGuard AI (iPSC monitoring)

Note: No demo video provided. Comment-response content embedded in the MCV text-box as "Summary of Changes Made" (shown in full below).

MCV text-box — full content De-facto Comment Response

Submitted on 28-Apr-2026 @ 03:33:04 from 110.164.228.206 · The text-box contains an extended version of the report (with content beyond what's in the attached docx) plus the "Summary of Changes Made" list that addresses each committee comment.

1. Executive Summary

The field of regenerative medicine is currently limited not by biological potential, but by the logistical constraints of manual laboratory management. CellGuard AI was conceptualized to solve the "Cost of Human Error" in induced Pluripotent Stem Cell (iPSC) research. In academic and clinical settings, the failure rate of cell batches remains high due to inconsistent visual analysis and a lack of 24/7 oversight. CellGuard AI introduces a specialized "AI Co Scientist" model that integrates computer vision with cloud-based alerting to provide autonomous surveillance. By automating the monitoring cycle, the system allows for an 80% reduction in manual labor while ensuring that critical culture failures such as oxygen level drops, heaters drying out, or unexpected overnight growth spikes spontaneous differentiation are caught during previously unstaffed hours.

3. Pain Points & ROI

Beyond general monitoring, validation pinpointed discrete incidents: oxygen drops stalling metabolism, heaters drying out causing lethal osmotic shifts, and faster growth speed overnight leading to over-confluency. Such errors and more cost approximately $2,500–$5,000 per batch. CellGuard AI delivers a clear Return on Investment (ROI), saving a mid-sized lab roughly $120,000 annually in reallocated salary costs.

4. Market Positioning & Competitor Analysis

Unlike equipment giants selling expensive closed hardware, CellGuard AI is a SaaS platform. Our unique value is the personalized AI feedback loop where the system learns the specific morphology nuances of a lab's unique cell lines using their available imaging materials. Unlike existing reactive incubator cameras that offer only basic image storage, CellGuard provides validated biological security through a software-first approach.

5. Technical Specifications & System Architecture

Front End & UI (Lovable Integration): The functional system dashboard was generated using the Lovable development platform. This interface serves as the primary touchpoint for researchers, offering real-time morphology scoring, confluency trend tracking, and a digital audit trail. Lab Edge Capture Hardware: incubator modules capturing high-frequency time-lapses. Cloud AI Backend: Data is preprocessed at the edge to reduce bandwidth load before being transmitted to a scalable cloud backend; the AI applies objective morphology scoring. Mobile Monitoring & Alerting: instant push notifications.

7. Strategic Roadmap to Commercialization

Phase 1: Launch within a selected research lab network to gather diverse iPSC datasets. Phase 2: Expansion into "Multi Line" monitoring, covering Mesenchymal Stem Cells (MSCs) and primary cells. Phase 3: Achieving the FDA Regulatory Mark — shift to "Virtual Monitoring," where software-only time-based simulations forecast in advance, allowing for proactive intervention.

Summary of Changes Made (de-facto Comment Response — text-box bottom)

Software-Only Pivot: I moved the "Revenue Streams" and "Cost Structure" on Slide 13 away from hardware sales. The startup is now positioned as a SaaS that uses whatever materials are already in the lab.
Technical Detail: I added the three specific "Pain Points" (Oxygen, Heaters, Growth Speed) to Slide 3 to satisfy the committee's request for technical clarity.
Financial Proof: I integrated the 1,612% ROI calculation onto Slide 8.
Competitive Edge: I updated Slide 10 to emphasize that no new hardware is required, which is your primary advantage over existing companies.
Future Vision: I added "Predictive Simulations" to Slide 14 to show that the technology is moving from just "watching" cells to "forecasting" their health.
Integrated Pain Points: Defined the specific technical failures (oxygen drops, heater failures, and overnight growth speed) the system solves.
Report: Parallel revisions were applied to the report for consistency (highlighted).

Acknowledgements (verbatim)

During the design, conceptualization, and technical documentation of CellGuard AI, the designer utilized AI assistants, including Gemini and Claude, for research synthesis and architectural drafting, alongside Lovable for system interface generation. While these tools acted as essential assistants in refining the business idea and generating the technical report, the designer retains full responsibility for the system's final design, its practical application, and the accuracy of the following report.

Missing / Non-compliant

❌ Demo Video — not submitted in any form.
❌ Validation evidence — no named users, slide 12 "Initial system feedback:" is empty; no validation summary file.
⚠ Comment Response Document — no separate file; content embedded in MCV text-box ("Summary of Changes Made").
⚠ Pitch deck format + length — PPTX (brief asks for PDF), 14 content slides vs 10–12 cap.

Presented in class

6878301130 + 6778310730

Kittitat Jaidee & Papawee Chennavasin Pair

Project 5 · NutriTrack AI (Enteral feeding)

Pitch Deckpdf
Validation Summarypdf
Final Reportpdf
Comment Response (Kittitat)pdf
Additional Report (Papawee)docx
Prototype link (.txt)txt
↗ Live MVP · nutritrack-thai.lovable.appweb

MCV text-box content Pointer + group composition

Submitted on 07-May-2026 @ 01:33:18 — Kittitat's text-box. Used to declare team composition + reusable links.

"Dear Aj.Pravee, I have attached the relevant files for final deliverables.

Our group have 2 students:

1. Papawee Chennavasin
2. Kittitat Jaidee

Our web link project is nutritrack-thai.lovable.app

PDF files were uploaded in MyCourseVille. And for Video Demo, due to large file size, I provided with link for download in Google Drive: drive.google.com/.../1dKNLl9ZrJ7iSNmA43BbjklZDEBRW5Es-/view

sincerely"

Demo video · 5:57 ✅

Missing / Non-compliant

⚠ Validation evidence under-quota — only 2 users (nurse + family caregiver); brief requires 2 per student → 4 for a 2-student team.
⚠ Pitch deck length — 14 content slides vs 10–12 cap.
⚠ Final report length — 10 pages covering 5 prior assignments; likely well over the 1000-word limit.

All 6 deliverables present; no missing files.

Presented in class

6778304030

Chayanee Sae-lim Solo · Not in feedback

Project 6 · OonJai Breast Partners

Note: Pitch deck exists from the 22-Apr pre-final upload (13 slides); not re-uploaded with the 04-May revised report. Demo video supplied via YouTube link (outside the MCV file slot).

MCV text-box content File pointer only

Submitted on 04-May-2026 @ 12:59:10 — text-box used only to point at the latest version of the revised report; no comment-response narrative in the text-box itself.

"The revised file is OonJai_Updated_Chayanee 6778304030_2.docx. Thank you ka."

Demo video · 3:15 ⚠ under 5 min

Missing / Non-compliant

⚠ Pitch Deck — pre-final version from 22-Apr (13 content slides — one over the 10–12 cap); not re-uploaded with the 04-May revised report, but file remains accessible.
❌ MVP URL — Lovable prototype mentioned in the report but no public link provided in the MCV slot.
⚠ Demo Video — provided via YouTube link, outside the MCV file slot; duration 3:15 is well under the 5-min floor.
⚠ Comment Response — embedded in the revised report, not a separate document. (Also: OonJai is not covered in Feedback.docx, so adaptation check is skipped.)

Pending in-class presentation

Students with no submission

Student ID	Name	Status
6878004830	Khalilullah Arsyi	No submission — empty MCV slot
6878005430	Khin Sandar Win	No submission — empty MCV slot
6878006030	Khin Htet Htet	No submission — not even in MCV submitter list

4 · Interactive Scoring Dashboard

Pick a project below to see its full grading. Every score has a short reason, a list of cited quotes (file + section/slide), and one concrete improvement. The AI commentary (impartial Claude Opus 4.7 verdict) sits at the bottom of each panel.

How scoring works: 5 criteria sum to the 40% course component — Problem-Solution Fit (10) · MVP (5) · Validation (5) · Business Viability (10) · Presentation & Pitch (10). Scoring scale: 90–100% Exceptional · 75–89% Solid · 60–74% Adequate · 40–59% Weak · 0–39% Inadequate.
* Asterisks on OonJai indicate provisional scores pending the in-class pitch.

Group	Project · Lead	P-S Fit /10	MVP /5	Validation /5	Business /10	Pitch /10	Total /40	%	Band
B	EV Optima AI · Chadaporn	8.5	4.5	4.0	8.5	8.0	33.5	83.75%	Solid → Exceptional
E	NutriTrack AI · Kittitat + Papawee	8.5	4.5	3.0	8.5	8.0	32.5	81.25%	Solid → Exceptional
A	Menopause Chatbot · Xin Cai	7.5	4.0	3.5	7.5	8.0	30.5	76.25%	Solid
F	OonJai Breast Partners · Chayanee Pending in-class pitch	7.5	4.5	4.0	8.0	6.0*	30.0*	75.0%*	Solid
C	MindCheck · Xiaoyan Li	6.5	4.0	2.0	7.5	6.0	26.0	65.0%	Adequate
D	CellGuard AI · Zinab	7.0	3.5	1.5	7.5	5.5	25.0	62.5%	Adequate
Non-submitters: Khalilullah Arsyi (6878004830), Khin Sandar Win (6878005430), Khin Htet Htet (6878006030) — 0 / 40 on the group-project component unless special circumstances apply.

↑ Quick-reference table · Click any project below for the full per-criterion breakdown with cited quotes.

Project 2 — EV Optima AI

Chadaporn Attakitbancha · 6878303430 · Solo

33.5 / 40

83.75% · Solid → Exceptional

Problem-Solution Fit /10i8.5 / 10

🎯 ProblemEV production is inefficient — researchers rely on trial-and-error across many culture conditions, with high cost, long timelines, and poor reproducibility. Three named user pains: choosing optimal conditions, interpreting multiplex datasets, ensuring reproducibility for clinical translation.

💡 SolutionEVOptima AI — a B2B predictive-analytics SaaS that ranks experimental conditions, flags biological risks, and visualises multi-parameter trade-offs (radar plots, cytokine heatmap, custom-weight sliders).

🔗 FitTight. Each of the three named pains has a corresponding feature: AI-driven ranking → "optimal conditions"; biological-interpretation layer → "complex datasets"; continuous-learning loop → "reproducibility". Customer insight is interview-validated but interview-only — no lab has yet used the platform on its own data.

EV production remains highly inefficient and inconsistent due to the complex interplay of cell culture conditions… Current optimization relies heavily on trial-and-error experimentation, resulting in high costs, long development timelines, and significant variability across laboratories.
Final_Report.docx · §1 Project Journey
EV researchers face three core challenges: 1. identifying optimal culture conditions, 2. interpreting complex datasets (e.g., cytokines, growth factors, mitochondrial markers), and 3. ensuring reproducibility and scalability for downstream clinical translation.
Final_Report.docx · §1
Too many experiments, not enough direction / High cost, time-consuming, low reproducibility / Difficult to interpret complex data
Pitch Deck p.3 — researcher interview quotes

Why 8.5 / 10 — Solid → ExceptionalLands in the upper-Solid band because the problem is real, well-articulated, AND validated through 2 researcher interviews — not just literature. The solution is engineered against the three named pain points one-to-one. Held below 9.0 because customer insight is interview-only; the platform has not yet been used by any lab on its own real data, which is the test of true problem-solution fit.

MVP /5i4.5 / 5

🛠️ DescriptionA B2B analytics dashboard: upload an experimental cytokine dataset → AI ranks the conditions, computes Anti-Inflammatory / Regenerative / Mitochondrial / EV-Functional scores, surfaces "best conditions", flags inflammatory-risk profiles, and lets the researcher re-weight cytokines via sliders. Includes radar plots, EV-score ranking bar chart, and a cytokine heatmap.

📊 StatusLive + public at ev-boost-insight.lovable.app. Demonstrated end-to-end in the demo video on a real experimental dataset (15 cytokines × 10+ culture conditions). Underlying logic today is rule-based scoring on uploaded data, not a trained predictive model.

AI Recommendation: Cytokine 4 · Best Conditions: Hypoxia · Key Drivers: Improved EV functional performance, Low inflammatory risk · EV Score 49 / 100
Pitch Deck p.6 — live dashboard screenshot
Anti-Inflammatory Score 33 · Regenerative Score 36 · Mitochondrial Score 32 · EV Functional Score 35 · Inflammatory Risk 0
ev-boost-insight.lovable.app — public MVP
Real experimental cytokine dataset shown on screen: Free / Free / Hypoxia / NPN / H2O2 / Cytokine-4 / IL-1β / TNF-α / IFN-γ / LPS rows × 15 cytokine columns including BDNF, EGF, FGF-2, HGF, IDO, IL-4, IL-6, IL-10, LIF, PDGF-BB, SDF-1α, TNF, VEGF-A.
Demo video 4:01 + EV_Optima_AI_Validation_Board_result.xlsx

Why 4.5 / 5Near-perfect. The MVP is live, public, exercises a real experimental dataset, and shows clean UX across multiple views (ranking · heatmap · radar · weight sliders). Held below 5.0 only because the AI is rule-based scoring on uploaded data rather than a trained predictive model — the team is honest about this, but a perfect score would require the predictive layer to actually work on new unseen data.

Validation Evidence /5i4.0 / 5

👥 Number2 — meets the brief's "2 potential users per student" minimum for a solo project.

🆔 Who2 EV researchers (working scientists, not students). Identities are anonymised in the deliverables; quotes are presented as collective.

📍 StatusPre-build interviews only — the platform has not yet been used by any user on their own data. The continuous-learning loop (User Input → AI Prediction → Experimental Validation → Data Re-upload → Model Refinement) is architecturally wired but not yet executed with a real user.

Interviewed 2 EV researchers — Validated key pain points in EV optimization. Identified key pain: Trial-and-error experimentation and data complexity. Built MVP platform (live demo).
Pitch Deck p.9 — Traction / Validation
A multi-source data acquisition and collaboration strategy was added, including internal data, academic partnerships, industry collaborations, and public EV databases, along with a continuous learning feedback loop.
Additional_Comment_Response_Document.docx · §1
Continuous Learning Loop: User Input → AI Prediction → Experimental Validation → Data Re-upload → Model Refinement
Final_Report.docx · §3.3

Why 4.0 / 5Solid because (a) N=2 meets the brief minimum, (b) pain points triangulated across the two researchers and re-stated verbatim on the deck, and (c) the response document explicitly cites validation-driven changes (the data-acquisition strategy was added in direct response to the Critical interim comment). Held below 4.5 because no user has yet used the prototype on real data — N=2 is the floor, not the ceiling.

Business Viability /10i8.5 / 10

💰 Revenue modelHybrid SaaS + usage. $30/mo platform access · $10/dataset analysis · custom enterprise tier. Average revenue/lab ≈ $80/mo assuming 5 analyses/lab/mo.

🎯 MarketTAM: global life-science R&D software (AI + lab informatics). SAM: biotech + academic labs doing cell-therapy / EV / cytokine research. SOM: EV + cell-therapy labs in Thailand / SE Asia.

📈 GTM & economicsConference channel (ISEV, ISCT) + B2B engagement + academic collaborations. Quantified economic benefit: ~$5,500 saving per optimisation cycle vs baseline ($8,000 → $2,500); 75% labour reduction, 70% cost reduction. 3-year revenue projection: $10K → $38K → $115K (10 → 40 → 120 labs).

Platform Access: $30 per month, providing access to the core AI platform and basic analytics features · Usage-Based Pricing: $10 per dataset analysis · Enterprise Tier: Custom pricing for advanced integrations
Final_Report.docx · §5.1 Revenue Model
Year 1: ~10 labs · ~$800/mo · ~$10,000/yr · Year 2: ~40 labs · ~$3,200/mo · ~$38,000/yr · Year 3: ~120 labs · ~$9,600/mo · ~$115,000/yr
Final_Report.docx · §5.3 + Pitch Deck p.13
Without EVOptima AI (Baseline): ~20 experimental conditions · 2 researchers × 2 weeks · ≈160 labor hours · Total ≈ $8,000 per cycle · With EVOptima AI: ~5 conditions · 1 researcher × 1 week · ≈40 labor hours · Total ≈ $2,500 per cycle · Net saving ≈ $5,500 per optimization cycle · Labor reduction ~75%, Cost reduction ~70%
Pitch Deck p.11 — Competitive Advantage

Why 8.5 / 10 — Solid → ExceptionalPricing concrete (not "TBD"), 3-year revenue projection mathematically grounded, and the with-vs-without cost comparison is the cohort's gold standard for quantified economic benefit (directly answering the committee's Critical interim comment). Held below 9.0 because 120 labs by Year 3 is an aggressive adoption curve given the relatively small global ISEV community — no sensitivity case provided.

Presentation & Pitch /10i8.0 / 10

📐 ClarityHigh. Slide 2 ("What are EVs / Why EVs Matter / The Core Challenge") is dense-info-done-right — three columns with clear visual hierarchy. Each section has a heading + bulleted key points.

📖 StorytellingStrong narrative arc: science intro → problem → solution → market → business → traction → competitive advantage → marketing → financials → vision. The deck escalates from "what" to "why fund us" smoothly.

🎤 PersuasionPersuasive — the $5,500/cycle saving + 75% labour reduction land hard. In-class pitch already delivered.

1. What are EVs? · 2. Why EVs Matter · 3. The Core Challenge — Same Cells × Different Conditions → Different EV Outcomes. High variability in EV yield and function · Multiplex data is complex and hard to interpret · No clear way to determine the best condition.
Pitch Deck p.2 — three-column framing
13 content slides vs the brief's 10–12 cap (one over). Demo video 4:01 vs the brief's 5–7 minute floor (one minute under).
Compliance check vs MDCU_3000747 brief p.3

Why 8.0 / 10 — SolidClarity and storytelling both Solid → Exceptional level. Held below 8.5 by two compliance gaps: deck runs 13 content slides (one over the 10–12 cap) and the demo video is 4:01 (one minute below the 5-min floor). Not a hit to the substance, but the rubric explicitly weights "clarity + storytelling + persuasion skills" equally — sloppy compliance signals to a panel that the team didn't follow brief.

AI Verdict · Claude Opus 4.7Fund — 6-month pilot

The most fundable submission in the cohort by a clear margin. The B2B SaaS angle is realistic, the economic math is concrete ($5,500 saving per cycle, 75% labour reduction), and the data-acquisition strategy directly answered the Critical interim concern. I would fund this for a 6-month pilot.

Fix first: the platform has never been used on real lab data — get one CDMO or academic EV lab onto a 4-week pilot before pricing it at $30/mo, because $30/mo is too cheap if it works and too expensive if it doesn't. No integrity concerns.

Project 5 — NutriTrack AI

Kittitat Jaidee (6878301130) + Papawee Chennavasin (6778310730) · Pair

32.5 / 40

81.25% · Solid → Exceptional

Problem-Solution Fit /10i8.5 / 10

🎯 ProblemCaregivers of tube-fed patients at home struggle with formula-specific dose math, tolerance titration, and waiting weeks for the next clinic visit to adjust feeds. Backed by peer-reviewed evidence: enteral patients have significantly higher 30/90-day readmission risk; top causes include access-device issues, GI symptoms from EN, and sodium imbalance from feed/flush errors.

💡 SolutionNutriTrack AI — a bilingual (TH/EN) clinical-decision web app that takes patient weight, formula brand, tolerance, and meal schedule, then computes a safe daily volume + step-up titration plan + preparation guidance + intake-vs-target tracking.

🔗 FitDirect and physician-validated. Founded by MDs across multiple specialties who observed the gap themselves — the "demand pull" framing is honest. Each named pain (calculation complexity, tolerance titration, hospital bottlenecks) has a corresponding feature.

Caregivers and patients often struggle to manage complex enteral feeding schedules at home due to limited gastrointestinal tolerance and the wide variability in nutritional formulas and caloric compositions among medical foods.
05_Final_report.pdf · Assignment 1 (Journey)
Patients with enteral nutrition have a significantly higher risk of 30- and 90-day readmission compared to those without. Top three causes: Issues with the enteral access device · GI symptoms from EN · Sodium imbalance, related to feed and flush doses.
Pitch Deck p.3 — Current Gastroenterology Reports 25(1), 61-68 (2023)
NutriTrack AI was founded in 2026 by Kittitat and Papawee… physicians from multiple medical specialties recognized a critical nutritional bottleneck in patient recovery.
05_Final_report.pdf · Assignment 1 — founder credibility

Why 8.5 / 10 — Solid → ExceptionalUpper-Solid because the problem is peer-reviewed, founder-credibility is real, AND the solution maps to the specific named gaps (titration, formula complexity, post-discharge bottleneck) with feature-level granularity. Held below 9.0 because customer insight is documented only after the MVP was built — pre-build interviews would strengthen the fit argument.

MVP /5i4.5 / 5

🛠️ DescriptionBilingual (Thai/English) clinical decision-support web app. End-to-end caregiver flow: Patient Info (sex/weight/height) → Formula Selection (Ensure / Peptamen / Blendera / Neomune with kcal/protein per scoop) → Nutrition Goal (kcal/kg/day, g/kg/day protein) → Current Tolerance → AI-computed daily volume, scoops-per-meal, water flush, fluid-loss adjustment, step-up titration over Day 1–3+ → Diary tab for actual-intake vs goal logging.

📊 StatusLive + public at nutritrack-thai.lovable.app. Highest-fidelity working flow in the cohort. Demo video 5:57 walks the full clinical path. Medical-food database is currently small — scaling SKU coverage is the named major engineering challenge.

Patient Information (Sex / Body Weight 70kg / Height 175cm / Ideal Body Weight 72kg) · Formula Selection (Ensure 1 kcal/ml 37 g/L, Peptamen, Blendera, Neomune) · Nutrition Goal (30 kcal/kg/day, 1.2 g/kg/day protein) · Current Tolerance (300/300/300/300 ml)
Pitch Deck p.13 — live screenshots
Output: Total 2741 ml · Preparation per Meal: Breakfast 13 scoops + water to 541 ml … · Step-Up Titration Plan: Day 1 1500 ml → Day 2 1700 ml → Day 3 1783 ml · Daily Water Plan +47 ml per meal · Feeding Diary 1800/1842 ml (98%) compliance tracking
Pitch Deck p.14 + demo video 5:57
Available in TH and EN (language toggle visible in app header)
nutritrack-thai.lovable.app — live public MVP

Why 4.5 / 5The MVP is the closest thing in the cohort to a real product — bilingual, end-to-end clinical flow, output explicit enough that a real caregiver could follow it. Held below 5.0 because the underlying medical-food database has only 4 SKUs (Ensure / Peptamen / Blendera / Neomune); Bangkok hospital coverage would need ~20–30 SKUs minimum. The team names this explicitly as their major challenge.

Validation Evidence /5i3.0 / 5

👥 Number2 — but the brief requires 2 per student, so for a 2-student pair this should be 4. Half of the required quota.

🆔 Who(1) Hospital Nurse in postoperative enteral feeding care (clinical-workflow expert), (2) Family caregiver of a patient on home tube feeding (end-user). Roles are stated; names anonymised.

📍 StatusHands-on prototype demo + structured feedback collection. Each validator named specific feature requests that the team then implemented — the clearest "we changed X because user said Y" pattern of any sub-5-score team.

1. Healthcare Provider Validation — Hospital Nurse in postoperative enteral feeding care. Status: Workflow Relevance & Practical Value Confirmed. Requested Features: Discharge Feeding Summary · Safety Alert System for diarrhea, vomiting, abdominal distension, or poor intake tolerance.
02_Validation_Summary.pdf · Validator 1
2. Caregiver Validation — Family caregiver of patient receiving home tube feeding. Status: Problem-Solution Fit Confirmed. Requested Features: Step-by-Step Preparation Guide · Reminder and Notification Tools · Daily Intake Log… submit information to the hospital system for follow-up.
02_Validation_Summary.pdf · Validator 2
Nurse-requested "Discharge Feeding Summary" + "Safety Alert System" became Slide 15 Workflow Integration (Step 2 Discharge onboarding, Step 6 Daily tracking with symptom flagging). Caregiver-requested "Step-by-Step Preparation Guide" became Slide 13–14 caregiver-centric tabs and Diary.
Cross-reference: 02_Validation_Summary.pdf → Pitch Deck pp.13–15

Why 3.0 / 5 — AdequateCapped at 3.0 because the brief is explicit: "2 potential users per 1 student", and a 2-student team needs 4 validators. The team delivered 2. The quality of the 2 interviews is otherwise 4.5-level work — both validators gave concrete feature requests that visibly drove the deck redesign. If 2 more validators were added, this would be the cohort's strongest validation score.

Business Viability /10i8.5 / 10

💰 Revenue modelMulti-stream. B2C: 14-day free trial + 99–199 THB/month subscription + 49 THB / 7-day pay-per-use. B2B: hospital licensing per active patient/month. Sponsorship: brand-neutral medical-food database partnership. Commission: 3–5% referral fee on in-app reorders.

🎯 MarketThai enteral-feeding-devices market projected $62.4M by 2025, CAGR 6.1% through 2033. ~55% distribution via hospital pharmacies. TAM/SAM/SOM defined qualitatively as Thailand → Bangkok hospitals + home-care → King Chulalongkorn Memorial Hospital pilot.

📈 GTM & differentiationKCMH pilot → Bangkok hospitals + home-care networks → LINE OA + caregiver groups + Thai tutorials + medical-food-distributor channels. Tubie competitor table is cohort-best: 7 dimensions × NutriTrack/Tubie/Benefit columns with explicit "moves from recording to clinical planning" positioning.

14-day free trial after discharge · B2C subscription 99–199 THB/month · Pay-per-use 49 THB / 7-day feeding plan · B2B licensing: hospital fee per active patient/month · Sponsorship: brand-neutral medical food database partnership
Pitch Deck p.9 — Business Model
NutriTrack AI vs Tubie — Core function: AI-based enteral feeding calculator & management plan vs Feeding schedule, logging, reminders → "Moves from 'recording' to 'clinical planning'". Daily titration: Generates day-by-day titration based on tolerance vs User adjusts manually → "Safer step-up feeding at home".
Pitch Deck p.11 — 7-row × 3-column comparison
NutriTrack takes a 3–5% referral fee on every order placed through the app's "Buy Now" button for home delivery… Brands pay for access to anonymized, aggregated "Tolerance & Adherence" reports.
Additional_report.docx (Papawee) — Strategic Partnerships
Thai enteral feeding devices market projected to reach $62.4 million in 2025, growing at a CAGR of 6.1% through 2033. Approximately 55% of enteral products distributed through hospital pharmacies.
Additional_report.docx — Market Opportunity

Why 8.5 / 10 — Solid → ExceptionalCohort gold standard for monetisation diversity (4 streams), market sizing (real Thai numbers from Grand View Research), and competitive positioning (the Tubie spec table is investor-grade). Held below 9.0 because TAM/SAM/SOM remains qualitative (nested boxes with text labels) despite committee asking for concrete numbers — the team added the Thai $62.4M figure to the report but never moved it onto Slide 8.

Presentation & Pitch /10i8.0 / 10

📐 ClarityHigh. Slide 15 7-step Workflow Integration is the cohort's best single infographic. Slide 11 Tubie comparison is textbook competitive-table format. Slides 13–14 product screenshots are annotated and self-explanatory.

📖 StorytellingStrongest demo video in the cohort. Opens with a tired caregiver at bedside writing "FEEDING PLAN? How much formula?" by hand — establishes problem viscerally — then walks the prototype, then closes with the "Why NutriTrack AI Matters" before/after impact infographic.

🎤 PersuasionPersuasive — quantified before/after impact (60%→100% nutritional goal attainment, fewer hospital visits) lands. In-class pitch already delivered.

Workflow Integration — From discharge planning to home monitoring and follow-up: 1. Hospital assessment · 2. Discharge onboarding · 3. Caregiver input · 4. AI feeding plan · 5. Home feeding · 6. Daily tracking · 7. Clinical review. → Reduces manual calculation errors · Saves caregiver time · Supports safer home enteral feeding.
Pitch Deck p.15 — cohort's strongest single graphic
Demo video 5:57 within 5–7 min spec. Caregiver scenario opening → full prototype walkthrough → "Why NutriTrack AI Matters" before/after impact infographic (Caregiver Confidence · Faster Nutrition Target Achievement 60%→100% · Fewer Unnecessary Hospital Visits).
Demo video — Kittitat_NutriTrack_video.mp4
14 content slides vs the 10–12 cap (2 over).
Compliance check vs MDCU_3000747 brief p.3

Why 8.0 / 10 — SolidStorytelling and persuasion both land at near-Exceptional level — the demo video is genuinely cinema-quality. Held below 8.5 by deck length: 14 content slides (2 over the 10–12 cap) suggests the team didn't ruthlessly cut, and Slide 6 "Key Features" only has 2 cards (Smart Titration + Real-time Adjusts) which is a stub.

AI Verdict · Claude Opus 4.7Fund — KCMH pilot tomorrow

The MVP is closer to a real product than anything else in the cohort — a bilingual interactive web app at nutritrack-thai.lovable.app that already walks through patient intake → formula choice → titration plan → diary tracking. The Tubie comparison is investor-grade. I would fund this for a KCMH pilot tomorrow.

Fix first: recruit 2 more validators (a clinical dietitian KOL and an oncology caregiver) to satisfy the brief's "2 users per student" requirement; and add concrete TAM/SAM/SOM numbers to slide 8 — the committee asked for this and the team only addressed it qualitatively. No integrity concerns.

Project 1 — Menopause Chatbot (MenoChat)

Xin Cai · 6878305730 · Solo

30.5 / 40

76.25% · Solid

Problem-Solution Fit /10i7.5 / 10

🎯 ProblemWomen in perimenopause / postmenopause (~1 billion globally by 2030) struggle to get reliable, plain-language menopause information. Existing sources are fragmented, inconsistent, sometimes commercially biased; brief clinical consultations don't allow personalised education; clinical guidelines are too technical for lay use.

💡 SolutionMenoChat — a RAG-based conversational chatbot grounded in international menopause guidelines (IMS, NAMS). Provides plain-language Q&A, suggested follow-up questions, boundary control for off-topic queries, and safety-oriented responses for risk situations (suicide / self-harm).

🔗 FitSolid. RAG architecture directly attacks the misinformation pain (responses cite the IMS knowledge base). Conversational format directly addresses the "too-technical" gap. Customer insight is thin — 2 peer users, not actual perimenopausal women.

Many women still struggle to obtain reliable and understandable menopause information [3,4]. Existing information sources are often fragmented, inconsistent in quality, and sometimes influenced by misinformation or commercial interests.
5_..._Project_Report_revised.pdf · §1 Introduction
With nearly one billion women expected to be in the menopausal age group globally, menopause is not only an individual health issue but also an important public health concern.
Final Report §1 — citing Panay et al. (2025) IMS
Fragmented Information — "Different sources, inconsistent messages" · Low Health Literacy — "Medical terms difficult to understand" · Misinformation — "Social media & non-verified sources" · Healthcare Constraints — "Limited time & personalized education"
Pitch Deck p.3 — citing McCartney 2022, BMJ

Why 7.5 / 10 — SolidSolidly in the Solid band: problem is unambiguously real and well-cited (Lancet 2024, IMS 2025, BMJ 2022); solution architecture (RAG over curated guidelines) directly attacks the named pain. Held below 8.0 because customer insight comes from peer users in their 20s–30s — not the actual 45–60 target demographic — so the "fit" half of "problem-solution fit" is asserted, not validated.

MVP /5i4.0 / 5

🛠️ DescriptionConversational web chatbot with RAG (Retrieval Augmented Generation) over a curated menopause knowledge base. Login + welcome with starter questions, free-form chat, suggested follow-ups, citation panel, off-topic boundary control, and a risky-question safety response with Thai mental-health / medical / police hotlines.

📊 StatusLive + public at menochat.lovable.app. All 3 demonstrated flows (medical Q&A, off-topic deflection, self-harm safety) shown in the demo video. The knowledge base is asserted to be IMS-grounded but the indexed source list isn't surfaced.

The chatbot prototype uses Retrieval Augmented Generation (RAG) architecture. Rather than relying only on the internal knowledge of a large language model, the system first retrieves relevant content from a curated knowledge base and then generates responses grounded in that evidence [7,8].
Final Report · §3 Technical Approach
Risky question "I want to hurt myself" → safety response card titled "We care about you" + Mental Health Hotline 1323 · Medical Emergency 1669 · Police Emergency 191
Pitch Deck p.16 — Main Features (safety flow)
Boundary control demonstrated in demo video — "I have a stomache" → "Thank you for your question! However, MenoChat is designed specifically to provide information about menopause and midlife women's health."
Demo video 5:49 · menochat.lovable.app

Why 4.0 / 5Solid → near-Exceptional. The MVP is live, all three demonstrated flows work end-to-end, and the safety response is the cohort's most responsible feature (clinical chatbots should never miss self-harm escalation). Held below 4.5 because the RAG knowledge-base contents are unverifiable from the deliverables — IMS / NAMS are asserted as sources but no indexed list of which guidelines are actually loaded.

Validation Evidence /5i3.5 / 5

👥 Number2 — meets the brief's "2 potential users per student" minimum for a solo project.

🆔 Who2 female peer users (anonymous). Not in target demographic — peers are likely 20s–30s, while the chatbot's stated audience is women 40–60 in perimenopause/postmenopause.

📍 StatusHands-on prototype + Likert + open-ended. Each user interacted with menochat.lovable.app and rated Usability, Clarity, Willingness-to-Use plus free-text. Visible "X→Y" iteration: User 2's "scope should be broader" drove Slide 9 women's-health-platform pivot.

Initial validation was conducted with two female peer users. Each participant interacted with the chatbot prototype and then completed a short questionnaire containing Likert-scale items and open-ended questions.
Final Report · §5 User Validation
User 1 — Usability 5/5, Clarity 4/4, Willingness 4/5. User 2 — Usability 5/5, Clarity 4/4, Willingness 3/5. Mean: Usability 5, Clarity 4, Willingness 3.5.
Pitch Deck p.17 — User Validation results
User 2 Feedback: "The tool's scope should be broader (not limited for menopause)" → drove Slide 9 "From Menopause Chatbot to Women's Health Platform" scope-expansion pivot. User 1: "Integration with hospital services would further enhance its usability" → drove Slide 9 "Telehealth / clinic referral" Expansion 2 phase.
2_User_Validation_Result.pdf · Pitch Deck p.17 → p.9

Why 3.5 / 5 — SolidSolid floor — N=2 meets brief minimum, structured method (Likert + open-ended), and the clearest "X→Y" iteration trail (User-2 quote → Slide-9 pivot). Held below 4.0 because peers ≠ target demographic — perimenopausal women have different concerns (HRT decisions, bone health, cardiovascular risk) that 20s–30s users can't validate.

Business Viability /10i7.5 / 10

💰 Revenue modelFour streams. B2C: freemium + 100 THB/month, 250 THB/quarter, 800 THB/year. B2B SaaS: clinic/hospital/telehealth licensing (fixed monthly/annual fee). Telehealth integration: setup fee + per-completed-consult commission. Referral & booking: pay-per-success.

🎯 Market~1 billion women globally in menopause by 2030 (Lancet 2024). Target segments: women aged 40–60, English-speaking initially, caregivers + family. No Thai-specific TAM/SAM/SOM numbers despite the local THB pricing.

📈 GTM & capture3-phase market entry: international hospitals + clinics → community outreach (expat networks, wellness workshops) → digital campaigns (Facebook, Instagram, expat groups). Critical answer to "what if patient goes direct to hospital?" → revenue stream #4 captures pay-per-success referral even when care is external.

1. B2C Subscription — Free tier · Premium 100 THB/month, 250 THB/quarter, or 800 THB/year. 2. B2B SaaS Subscription — Clinics, hospitals, and telehealth platforms pay a fixed fee. 3. Telehealth Service Integration — Integration fee + service commission. 4. Referral and Booking Partnership — Pay-per-success referral model.
Pitch Deck p.8 — Revenue streams & pricing strategy
Capture value even when care is delivered by external hospitals or clinics.
Pitch Deck p.8 stream #4 — direct answer to committee's "what if patient goes direct" Critical comment
The proposed business model combines both B2C and B2B revenue pathways… freemium model, offering free basic access with limited usage while premium subscriptions (100 THB/month, with quarterly and annual discounts) provide unlimited access, personalization, saved history, and extended educational content.
Final Report §6 Business Model

Why 7.5 / 10 — SolidSolid because revenue diversification answers the committee's Critical "what if patient goes direct" question explicitly via stream #4, and pricing is concrete (THB, not "TBD"). Held below 8.0 because the addressable market is sized only globally — "~1 billion women" is a pretty number but no Thai patient count, no SOM, and no conversion-rate assumption to ground the THB pricing.

Presentation & Pitch /10i8.0 / 10

📐 ClarityHigh. Consistent peach/teal/yellow palette, clean iconography, one-idea-per-slide layout. Within slide-count + backup limits.

📖 StorytellingStructured arc: title → global stats → 4 named pain blocks → market → solution (6 features) → Human-in-the-Loop Care Journey (P-P-T diagram) → business model → revenue streams → scope-expansion roadmap → competition → impact → references + 4 backup slides showing prototype screens.

🎤 PersuasionPersuasive. Demo video 5:49 walks the live chatbot through medical Q&A, off-topic deflection, and the suicide-safety response — the last lands emotionally. In-class pitch already delivered.

11 content slides + 4 backup slides — within the brief's 10–12 / ≤5 spec.
Compliance check vs MDCU_3000747 brief p.3
Demo video 5:49 — within the brief's 5–7 minute spec. Opens with "Current Gaps in Menopause Education" → solution overview → live MenoChat walkthrough → off-topic deflection → risky-question safety response.
Demo video file 3_Menopause_Chatbot_Video.mp4
Slide deck narrative: title → problem context (global stats) → pain points (4 named) → market opportunity → solution (6 features) → Human-in-the-Loop Care Journey (P-P-T diagram) → business model → revenue streams → scope-expansion roadmap → competitive differentiation → impact → references.
Pitch Deck pp.1–12

Why 8.0 / 10 — SolidCleanest deck design in the cohort + compliant slide/video counts + the only deck where the safety flow is included as a feature, not a footnote. Held below 8.5 because Slide 11 "Impact" reads as marketing soft-tissue (Greater Confidence · Greater Awareness · Health Equity) without measurable target metrics — the team could and should commit to specific impact numbers.

AI Verdict · Claude Opus 4.7Conditional — not fundable today

This is the cohort's best example of revision discipline — every committee comment was answered with a specific deck page and report section, and the answers are substantive (genuine reframing of menopause as an early-launch use case, not a final scope). I would not fund this venture today — the addressable market is too narrow without the broader women's-health expansion actually built, and the validation pool was 2 peers who are not in the target demographic.

Fix first: validate with 5+ actual perimenopausal women on factual accuracy of chatbot answers, not just usability. No integrity concerns — the AI-use declaration is included and references are properly cited.

Project 6 — OonJai Breast Partners Pending in-class pitch

Chayanee Sae-lim · 6778304030 · Solo · Not in Feedback.docx

30.0* / 40

75.0%* · Solid

Problem-Solution Fit /10i7.5 / 10

🎯 ProblemBreast cancer patients face anxiety and confusion across treatment pathways — diagnosis decoding, surgery/reconstruction choices, hormonal/targeted-therapy decisions, survivorship. Clinical encounters don't allow personalised education; pharma-influenced content is unreliable. Pre-final deck Slide 2 names three pains: lack of sufficient knowledge, misinformation prevalence, limited doctor time.

💡 SolutionOonJai Breast Partners — a dual-purpose platform: (1) AI chatbot delivering NCCN-grounded personalised education via LINE Official Account; (2) longitudinal QOL/survival registry generating real-world evidence. Founded by a practicing breast surgeon at KCMH.

🔗 FitStrong, with founder credibility as a multiplier. The NCCN-grounded knowledge base attacks "misinformation"; LINE-native delivery (>50M Thai users) attacks "access"; the registry component answers the unstated downstream question "how do we know this works?". Validated by 2 clinical reviewers.

Breast cancer remains one of the most prevalent malignancies worldwide, creating substantial challenges in patient understanding, treatment adherence, and long-term survivorship care.
OonJai_Updated_Chayanee.docx · §1 Introduction
Lack of Sufficient Knowledge for Patients · Misinformation · Limited Doctor Time
Pre-final Pitch Deck p.2 — three named pains
OonJai Breast Partners is a dual-purpose digital health platform that (1) delivers personalized, guideline-aligned breast cancer education through an AI-powered chatbot integrated with LINE Official Account, and (2) generates structured real-world evidence through longitudinal quality-of-life (QOL) and survival outcome registries.
OonJai_Updated_Chayanee.docx · §2
Chayanee Sae-lim, MD · General Surgeon, Fellow in Breast Surgery, KCMH · Ph.D. student, Clinical Science, CU.
Pre-final Pitch Deck p.1 — author credential

Why 7.5 / 10 — SolidSolid: problem is real, founder is the right founder, dual-purpose architecture is genuinely innovative. Held below 8.0 because the consent/IRB risk of the registry side isn't addressed in the deliverables — the dual-purpose pitch can also be a dual-purpose risk if patients sense data extraction.

MVP /5i4.5 / 5

🛠️ DescriptionNCCN-grounded oncology web app. Intake: TNM staging · ER/PR/HER2 subtype · surgery + reconstruction · axillary surgery · family + BRCA1/2 history · PDPA consent. "Stage II — In active treatment" dashboard with an AI guide answering from NCCN Breast Cancer v.2.2026. Monthly PRO questionnaire (nausea / fatigue / skin / numbness, 5-pt Likert). Follow-up logger with recurrence categories (NED / Local / Regional / Distant).

📊 StatusBuilt — Lovable + LINE OA integration described. Demonstrated end-to-end in the YouTube demo (3:15). No public URL surfaced in MCV slot. Validated by 2 clinical reviewers (per §7.2 of the report).

Comprehensive oncology intake: Clinical stage (TNM with T1–T4, N0–N3, M0/M1) · Tumor subtype (ER%/PR%/HER2 status) · Breast surgery (BCS / Total mastectomy / Nipple-sparing) · Breast reconstruction (Implant / TRAM-DIEP / LD flap) · Axillary surgery (SLNB / ALND) · Family & genetic history (first-degree relatives + BRCA1/2 status) · PDPA consent.
YouTube demo video 3:15 — intake form walkthrough
Ask Guide → "OonJai Guide" answering: "Articles draw answers from NCCN Breast Cancer v.2.2026". Suggested questions: What does HR+/HER2- mean for me? · How do I prepare for radiation after lumpectomy? · CDK4/6 inhibitor side effects? · SLNB vs ALND?
Demo — Stage II treatment dashboard
Monthly Check-in PRO questionnaire (nausea / fatigue / skin or hair changes / numbness or tingling, 5-point Likert "Not at all → Very much"). Follow-up status form: NED · Local recurrence · Regional recurrence · Distant metastasis.
Demo — PRO + Follow-up flows
Lovable-built prototype validated with 2 clinical reviewers; chatbot rated effective.
OonJai_Updated_Chayanee.docx · §7.2 Feasibility Assessment

Why 4.5 / 5The most clinically sophisticated MVP in the cohort — real oncology taxonomy (TNM × subtype × surgery × BRCA), structured PRO collection, recurrence tracking. This is the only MVP that looks like real clinical-research infrastructure rather than a class deliverable. Held below 5.0 because the public URL isn't surfaced in the MCV slot — a reviewer can't independently exercise the app.

Validation Evidence /5i4.0 / 5

👥 Number2 — meets the brief's minimum for a solo project.

🆔 Who2 clinical reviewers (oncology / breast-surgery context). Not actual breast-cancer patients — both are clinicians evaluating on behalf of the patient population.

📍 StatusHands-on testing with implemented changes. Cleanest "we changed X because user Y said Z" trail in the cohort. Reviewer 1 → centralised menu + login/logout (Figure 2 → 3). Reviewer 2 → BRCA-aware content driven by family-history field (Figures 4 → 5).

Reviewer 1: Suggested adding a centralized menu tab to improve navigation and allow easier access to all platform functions. Recommended implementing login and logout functions to enhance patient data confidentiality. Noted that the chatbot performs well and provides effective responses.
OonJai_Updated_Chayanee.docx · §9 Validation Evidence
Reviewer 2: Recommended including family history of breast, ovarian, and pancreatic cancer to enable personalized recommendations, particularly regarding genetic testing.
OonJai_Updated_Chayanee.docx · §9
Implementation evidence: Figure 2 (Previous version) → Figure 3 (Current version with centralized menu + login/logout). Figure 4 (registration including family history of breast, ovarian, and pancreatic cancer) → Figure 5 (Customized content regarding BRCA mutation and multigene assays).
OonJai_Updated_Chayanee.docx · §9 — before/after figures

Why 4.0 / 5Strongest validation narrative in the cohort despite small N — the before/after figures make the "X→Y" chain tangible. Held below 4.5 because both reviewers are clinicians; the platform's primary users are breast-cancer patients themselves, who weren't part of the validation round.

Business Viability /10i8.0 / 10

💰 Revenue modelPhase 1 (research): grant-funded (HSRI + KCMH research budget) — non-commercial. Phase 2 (venture): "Hidden Revenue" — sponsors pay (pharma content partnerships, non-promotional, independently reviewed), users free. Five partner categories.

🎯 Market~17,000 new breast cancer cases/year in Thailand (NCCN 2023). LINE OA reach >50M Thai users (no app-download friction). Competitive matrix vs Outcomes4Me (US, AI-driven cancer education), PatientsLikeMe (crowdsourced patient data), and Belong.Life (global cancer community).

📈 GTM & governanceIndependent medical board + transparent content labelling addresses the pharma-bias risk. Phased rollout (research → venture) matches the clinical-research product life cycle better than pure venture would. Commercialisation pipeline: Development → Testing → Regulatory → Launch → Growth.

Phase 1: Research Phase (Year 1–2) — funded via HSRI grant + KCMH research budget. Phase 2: Venture Phase (Year 3+) — via B2B/B2C revenue.
OonJai_Updated_Chayanee.docx · §3 Phased Business Model
Hidden Revenue — Sponsors pay · Users free · Know more about products. Freemium-influenced ecosystem where patients access services free of charge, while pharmaceutical sponsors support operational sustainability through content partnerships (non-promotional, independently reviewed).
Pre-final Pitch Deck p.7 + §4 + §5
Content Governance Structure — independent medical board + transparent content labelling. Five partner categories with defined collaboration structures.
OonJai_Updated_Chayanee.docx · §5.1
~17,000 new breast cancer cases/year in Thailand (NCCN 2023); LINE OA reach >50M Thai users. Competitive matrix vs Outcomes4Me, PatientsLikeMe, and Belong.Life.
OonJai_Updated_Chayanee.docx · §6 + §7.2

Why 8.0 / 10 — SolidSolid: phased model is the smartest in the cohort for a clinical-research product, governance proactively addresses pharma-bias risk, and competitive positioning is well-scoped against three real global comparables. Held below 8.5 because Thai TAM ~17K cases/year is small — regional or other-cancer expansion mentioned but not quantified.

Presentation & Pitch /10i6.0* / 10

📐 ClarityPre-final deck (13 slides, one over the 10–12 cap) is reasonably clear — Problem (3 pains) → Solution diagram → Personas → Stakeholders → Value Proposition → Hidden Revenue → Competitors → Competitive Advantage → Commercialization Pipeline. Slide design is functional rather than visually distinctive.

📖 StorytellingCoherent narrative arc but uneven density. The "Hidden Revenue" framing (Slide 7) is the most distinctive concept. YouTube demo (3:15, well under 5-min floor) is the weakest video deliverable in the cohort despite the strongest underlying product.

🎤 PersuasionPending — in-class pitch not yet delivered. The deck + demo can carry persuasion partially, but the brief weights "persuasion skills" as a live-delivery component too. Score will re-evaluate after the live pitch.

13 content slides covering: Problem Identification (3 pains) · Solution (AI-driven · Cancer registry · Personalized contents · Survival & QOL follow-up · Chatbot support) · Customer Personas · Stakeholder Network · Value Proposition · Hidden Revenue · Competitors (Outcomes4Me) · Competitive Advantage · Commercialization Pipeline.
Pre-final Pitch Deck pp.1–13
Latest MCV submission slot contains only OonJai_Updated_Chayanee.docx. Earlier Slide_Project_Chayanee.pdf from 22-Apr exists but was superseded in the slot — file remains accessible at the S3 URL.
Slide_Project_Chayanee.pdf — pre-final, accessible
Demo video supplied via YouTube — 3:15 duration, well under the 5–7 minute floor. In-class pitch not yet delivered → criterion provisional until live presentation.
YouTube demo · vjzJEt3xXGE

Why 6.0* / 10 — Adequate → Solid (provisional)Pre-final pitch deck exists and is reasonably structured, so this is no longer the worst-of-cohort Presentation. Held at 6.0* because: (a) deck wasn't re-uploaded with the 04-May final submission (signals lower priority on presentation polish); (b) demo video is 3:15 — shortest in cohort, well under floor; (c) live in-class pitch is still pending. Asterisk = provisional — will revise after live pitch.

AI Verdict · Claude Opus 4.7Provisional — fundable potential

Clinically the most credible founder in the cohort (a breast surgeon MD), and the dual-purpose architecture (patient education + RWE registry) is the most original idea submitted. The YouTube demo video supplied after the initial review (3:15, NCCN-grounded staging intake + AI guide + monthly PRO + recurrence-status logging) shows the underlying MVP is genuinely sophisticated — a real clinical-research infrastructure prototype, not a class project. That said, the MyCourseVille submission slot is structurally incomplete — no pitch deck, no demo link in the slot itself, video posted to a personal YouTube.

Fix first: re-export the earlier slides as PDF and add the YouTube + Lovable links to the MCV submission so the deliverables match what was actually built. No integrity concerns — and the validation narrative (Figure 2/3/4/5 before-after) is the cleanest "we changed X because user Y said Z" evidence in the cohort.

Project 3 — MindCheck (Mental Health Triage)

Xiaoyan Li · 6678313530 · Solo

26.0 / 40

65.0% · Adequate

Problem-Solution Fit /10i6.5 / 10

🎯 ProblemThai university students face rising mental-health burden — medical students show depression 9.3–30.5%, stress 61.4%, suicidal ideation 12.8%. Cultural stigma (fear of losing face) blocks help-seeking; mental-health resources are scarce; ~20% of Thais experience mental-health challenges but only 23% seek professional help.

💡 SolutionMindCheck — a self-administered triage chatbot using validated screeners (PHQ-9 + GAD-7) that scores users into Mild / Moderate / Severe risk, then routes them to tiered care: self-help → online counseling → direct hospital referral. AI emotional-journal feature for ongoing engagement.

🔗 FitReasonable fit on paper — clinical screeners directly address the "how do I know if I need help?" question; tiered routing addresses the "where do I go?" question. Customer insight is largely literature-review (10+ academic refs) rather than primary student interviews.

Medical student study: Depression 9.3%–30.5% · Stress 61.4% · Suicidal ideation 12.8%. (Chiddaycha & Wainipitapong, 2021, Health Science Reports 4(4), e416)
Pitch Deck slide 3 — 10 academic references
Cultural Stigma (Fear of Losing Face / Being Judged): mental illness is still stigmatized in Thai society. Cultural stigma prevents students from seeking help due to fear of judgment or losing face.
Pitch Deck slide 3 — Pitakchinnapong 2019, Sojindamanee 2023
Approximately 20% of Thailand's population experience mental health challenges, only 23% seek professional help.
Pitch Deck slide 6 — Rhein & Nanni 2022, SAGE Open 12(4)

Why 6.5 / 10 — Adequate → SolidMid-Adequate to lower-Solid: the problem is unambiguously real (10+ peer-reviewed citations) and the screener-based solution is clinically sound (PHQ-9/GAD-7 are gold standard). Held below 7.0 because customer insight is exclusively literature-derived — no Thai student was interviewed before or during MVP design, so "fit" is theoretical rather than user-tested.

MVP /5i4.0 / 5

🛠️ DescriptionSelf-administered triage web app. Full PHQ-9 (depression, including Q9 suicidal-ideation screen) + GAD-7 (anxiety) walkthrough → scored end-states with risk stratification (Minimal-to-Mild / Moderate / Severe / Emergency) → tiered referral list (Crisis Support · Find Mental Health Services · Bangkok Hospital · Siriraj Hospital).

📊 StatusLive + public at mindlink-buddy.lovable.app. Video review revealed the prototype is substantially more complete than the deck conveys — both green "minimal-to-mild" and red "It's an emergency" end-states are functional.

Demo video shows the actual working flow: PHQ-9 Q9 displayed verbatim — "Thoughts that you would be better off dead, or thoughts of hurting yourself" — followed by 4-option Likert (Not at all → Nearly every day).
Demo video 4:27 — PHQ-9 walkthrough · mindlink-buddy.lovable.app
Two scored end-states displayed: GREEN "Your results suggest minimal to mild symptoms" PHQ-9=7 / GAD-7=7 with Support resources, AND RED "It's an emergency" PHQ-9=17 / GAD-7=15 with Crisis Support CTA. Tiered referrals: Crisis Support · Find Mental Health Services · Bangkok Hospital · Siriraj Hospital.
Demo video frames — risk stratification + emergency flow
Decision pathway: Assessment (PHQ-9/GAD-7) → Score → Risk (Mild 0–9 · Moderate 10–14 · Severe 15+) → Action (Mild: self-help · Moderate: online counseling · Severe: direct referral to hospital & specialists).
Pitch Deck slide 10 — Competition slide

Why 4.0 / 5Solid → near-Exceptional: the working prototype handles both gold-standard screeners with risk-tiered routing and a real emergency-escalation pathway with Thai hotlines. Held below 4.5 because the deck doesn't show any of this — the team massively underclaims their own MVP, which the rubric implicitly punishes through the Presentation criterion's clarity dimension.

Validation Evidence /5i2.0 / 5

👥 Number0 actual — only a future "PILOT PLAN" stating "at least >2" users. Brief requires 2 users actually tested for a solo project.

🆔 WhoNo named validators. The deck mentions an intended pilot at "1 University pilot site, 100–300 users", but neither identifies any specific student tester nor presents any user-derived data.

📍 StatusFuture plan, not execution. The demo video is a developer-led self-walkthrough — no user on screen, no quoted feedback, no "we changed X because user said Y" pattern.

PILOT PLAN: 1 University pilot site · 100–300 users internal testing (at least >2). EXPECTED METRICS: ≥30% complete assessment · ≥10% convert to counseling recommendation.
Pitch Deck slide 8 — Traction/Validation (future tense)
Slide 9 has section headers "Results & Scoring" and "User Feedback & Action Plans" but no actual user data, names, or quotations extracted from the content.
Pitch Deck slide 9 — empty content under headers
Demo video is developer-led self-walkthrough of the system — no user appears on screen, no quoted feedback, no "we changed X because user said Y" pattern.
Demo video 4:27 — system demo, not user testing

Why 2.0 / 5 — WeakWeak band because the rubric is explicit: "Did they conduct user testing or experiment with the user using a prototype?" The answer here is no — only a future plan. Not 1.5 (Inadequate) because the prototype itself is functional and a credible validation could be run on it tomorrow; the gap is execution discipline, not capability.

Business Viability /10i7.5 / 10

💰 Revenue modelCleanest chatbot-team revenue capture. B2B (primary): Pay-per-referral ฿100–300 per confirmed visit + SaaS ฿1,000–5,000/clinic/mo. B2B (secondary): Universities pay custom pricing for data + prevention programs. B2C (optional): Premium ฿49–149/mo for students. Direct answer to "what if patient goes direct?": clinic SaaS still captures value.

🎯 MarketThai students aged 18–25 (high-risk mental-health group). 23% of Thais experiencing mental-health issues seek help — a 4× gap implies large addressable demand. No concrete TAM/SAM/SOM numbers on the deck.

📈 GTM & captureThree channels: (1) Hospital integration — clinic partners drive supply + credibility. (2) University channel — institutional B2B + campus ambassadors. (3) Digital acquisition — TikTok / social media + free screening tools as lead-gen. Funnel logic: First Touchpoint → Decision Influence → Controlled Pathway → Revenue Capture.

Pay-per-Referral: ฿100 – ฿300 per confirmed visit · Monthly SaaS Subscription: ฿1,000 – ฿5,000 per clinic · University Plan: custom pricing · Premium (Students): ฿49 – ฿149 per month (optional).
Pitch Deck slide 7 — Pricing Overview
Customer Segments — Primary: Clinics / Therapy Centers (Pay for high-intent, pre-screened patients) · Secondary: Universities / Institutions (Pay for data, insights, and prevention programs) · End Users: Students (Drive engagement and adoption — low cost).
Pitch Deck slide 7 — B2B-primary clarification
Revenue Capture Logic: First Touchpoint → Decision Influence → Controlled Pathway → Revenue Capture (We earn from referrals and subscriptions at scale).
Pitch Deck slide 7 — funnel logic

Why 7.5 / 10 — SolidSolid: revenue capture explicitly answers the committee's Critical "what if patient goes direct" interim comment with the SaaS layer (not just referrals); customer segments are correctly tiered (clinics primary, universities secondary, students free users). Held below 8.0 because the "Universities pay" assumption is unvalidated — Thai universities have historically resisted paying mental-health vendor tools.

Presentation & Pitch /10i6.0 / 10

📐 ClarityMixed. Committee flagged "slides too messy" at interim — partially resolved but Slide 7 still packs business-model + revenue-capture-logic + pricing-overview + customer-segments + core-statement onto one frame. Format non-compliant: PPTX submitted, brief specifies PDF.

📖 StorytellingFunctional but uneven. Problem framing (Slide 3) is strong with 10+ refs; competitive positioning (Slide 10 "We solve 'Do I need therapy?'") is sharply written. But the working MVP — the team's strongest asset — is barely visible on the deck.

🎤 PersuasionIn-class pitch already delivered. Demo video 4:27 is under the 5-min spec floor. Persuasion would land better if video screenshots were on the deck.

Slides are too messy with too many texts. You should review the detail in the Visual Aids lecture.
Feedback.docx · Project 3 MindCheck — interim committee comment
Slide 7 packs the entire business model + revenue capture logic + pricing overview + customer segments + core statement onto one slide — at least 5 distinct concept blocks competing for attention.
Pitch Deck slide 7 — density issue
Deck submitted as PPTX (brief specifies PDF). 12 slides total / 11 content slides — within count but format non-compliant. Demo video 4:27 — under the 5-min spec floor.
Compliance check vs brief p.3

Why 6.0 / 10 — AdequateMid-Adequate band: the committee specifically called out density at interim and the team partially addressed it but Slide 7 remains a problem. Format compliance (PPTX vs required PDF) + video under-duration + working MVP not surfaced on the deck = 3 separate clarity gaps that reinforce each other. The substantive content is decent — execution presentation is what holds the score back.

AI Verdict · Claude Opus 4.7Not yet — close validation gap first

The business model is the cleanest of the chatbot teams and the problem is unambiguously real. The video review (after the initial scoring) revealed that the prototype is actually further along than the deck conveys — full PHQ-9 incl. suicidal-ideation Q9, GAD-7, two scored end-states including an emergency-red crisis-support flow, and a referral list with Bangkok and Siriraj hospitals. But the deliverable package is still incomplete — no Final Report file, no Comment Response Document, and no real-user testing (the video is developer-led system demo). I would not fund this until the validation gap is closed.

Fix first: run the validation — even 3–5 university students with the live Lovable prototype, captured in a 1-page summary — and surface video screenshots into the deck. Integrity flag — minor: committee asked to "verify AI-generated content before submission" and this comment has no explicit response.

Project 4 — CellGuard AI (iPSC Monitoring)

Zinab Idris Omer · 6871004530 · Solo

25.0 / 40

62.5% · Adequate

Problem-Solution Fit /10i7.0 / 10

🎯 ProblemiPSC research has an "overnight blind spot" (6 PM – 8 AM) when labs are unstaffed but cell cultures can fail catastrophically. Concrete failure modes: oxygen drops, heaters drying out, faster overnight growth → over-confluency. Each failed batch costs ~$2,500–$5,000. Researchers spend ~40% of working hours on manual incubator checks.

💡 SolutionCellGuard AI — incubator camera modules + cloud-based AI morphology scoring + mobile push alerts. The "AI Co-Scientist" framing: 24/7 autonomous monitoring, objective morphology scoring (no inter-researcher variability), digital audit trail. Pivoted post-feedback to SaaS-only on existing lab cameras.

🔗 FitStrong on paper — the only project in the cohort that genuinely fits the brief's "hardware-integrated system" requirement. Three named failure modes (oxygen / heater / growth) each map to alert triggers. But customer insight is a single composite persona ("Dr. Sarah") rather than real iPSC researchers — the "40% of working hours" claim is asserted, not validated.

The field of regenerative medicine is currently limited not by biological potential, but by the logistical constraints of manual laboratory management.
Final_Report.docx · Executive Summary
Dr. Sarah represents the thousands of scientists who spend roughly 40% of their working hours performing repetitive, manual incubator checks. The "Emotional Satisfaction Curve" identified in the initial research highlights a massive "Pain Zone" during the incubation and manual check phases.
Final_Report.docx · §2 The Designer's Journey
Beyond general monitoring, validation pinpointed discrete incidents: oxygen drops stalling metabolism, heaters drying out causing lethal osmotic shifts, and faster growth speed overnight leading to over-confluency. Such errors and more cost approximately $2,500–$5,000 per batch.
Final_Report.docx · Pain Points & ROI section

Why 7.0 / 10 — SolidLower-Solid: the problem framing is concrete (named failure modes, dollar costs, time percentages) AND this is the only cohort project that meets the brief's "hardware-integrated system" requirement. Held at 7.0 because the customer insight is a single composite "Dr. Sarah" persona — no actual iPSC researcher was interviewed to validate that her experience is representative.

MVP /5i3.5 / 5

🛠️ DescriptionConcept architecture: incubator cameras → cloud AI morphology scoring → mobile push alerts for oxygen drops, heater failures, overnight growth anomalies. Dashboard generated via Lovable. Six value pillars in the slides (24/7 monitoring, instant alerts, objective scoring, digital audit trail, labor efficiency, stage troubleshooting).

📊 StatusConceptual + early Lovable dashboard. Slide 4 architecture is clean. Slides 5–9 in the deck contain only fragmentary placeholder text ("Connect the visual input device to the Ipscs (system)", "Tabs for navigation") suggesting the dashboard UI exists but wasn't documented at investor-deck quality. Lovable URL referenced but not surfaced for reviewer use.

AI-Assisted Monitoring flow: Seed Cells → Incubate (37°C) → AI Camera Capture → Analysis → Auto Alert & Log.
Pitch Deck slide 4 — system architecture
Slides 5–9 contain fragmentary placeholder text only: "Connect the visual input device to the Ipscs (system)" · "Tabs for navigation" · "Type of systems" · "Visual input" · "factors input" · "Model 2 will define the error" · "After optimizing the focus and zoom on the microscope you can click for evaluation" · "Fill all necessary information".
Pitch Deck slides 5–9 — draft-state content
Lovable URL referenced: lovable.dev/projects/eec6cf0f-9733-4354-abc8-05b8333c2554
Final_Report.docx · References

Why 3.5 / 5 — Adequate → SolidMid-band: the architecture is sound and an alert system targeting three named failure modes is feasible, but the deck doesn't document the working dashboard at investor quality — Slides 5–9 read as draft notes. The Lovable URL exists but isn't surfaced; a reviewer can't independently verify what was built.

Validation Evidence /5i1.5 / 5

👥 Number0 — brief requires 2 users for a solo project. None delivered.

🆔 WhoNo named users. The deliverable uses a single composite persona "Dr. Sarah" as a stand-in for real iPSC researchers.

📍 StatusAbsent. Slide 12 header reads "Initial system feedback:" but the body is empty in slide extraction. No validation summary file. No demo video. The Final Report's "validation research" section cites only market-size figures, not user testing. Self-citation problem: primary reference is "Idris, Z. (2026). [Unpublished manuscript]" — circular.

Initial system feedback: [empty — body text not present in slide extraction]
Pitch Deck slide 12 — title header only, no body content
Validation research was conducted to quantify the market need. The global iPSC market is expanding rapidly, with a Total Addressable Market (TAM) reaching up to $25.7B.
Final_Report.docx · §2 — cites only TAM figures, not user testing
Idris, Z. (2026). CellGuard AI: Precision automation for advanced cell therapy [Unpublished manuscript]. — primary citation in the references is the team's own unpublished writeup, which is circular.
Final_Report.docx · References §1 (self-citation)
No demo video submitted in any form — entire user-evidence vector for the Validation criterion is missing.
MCV submission inventory · 6871004530 (no video file)

Why 1.5 / 5 — Weak → InadequateWorst validation score in the cohort, just above Inadequate. The rubric asks two questions ("Did they conduct user testing?" and "Do they iterate on user feedback?"); the answer to both is no. Not zero because the team is honest about the gap in the AI-use acknowledgment, and the failure modes (oxygen / heater / growth) could be argued to come from secondary literature even if not user-validated.

Business Viability /10i7.5 / 10

💰 Revenue modelMulti-tier SaaS post-pivot. Implementation fee: $1,500–$3,000. Recurring subscription: $200–$500 / lab / month. Enterprise annual: $15K–$50K/year. Professional services: $5K–$20K per project. ROI math: $125K annual savings vs $7,300 annual cost = 1,612% Y1 ROI, < 1 month payback.

🎯 MarketGlobal iPSC market TAM ~$25.7B (BioSpace). Customer segments: academic iPSC core facilities, biotech cell-therapy companies, pharma R&D, hospital translational research. Channels: direct sales to labs, biotech conferences, academic publication network.

📈 GTM & positioningSoftware-only pivot per committee feedback — the right move (positioned as SaaS that uses whatever materials are already in the lab, vs equipment giants selling closed hardware). 4-phase roadmap: MVP launch (research lab network) → Beta testing → Commercial release → FDA Regulatory Mark.

Annual Savings: $125,000 · Annual Cost: $7,300 · 1,612% First Year ROI · Payback Period: < 1 Month · Net Annual Value: $117,700 per lab · Labor Efficiency Gain: 16 hours/week per researcher.
Pitch Deck slides 8 + 14 — ROI math
Revenue Streams: Implementation & Integration Fee ($1,500–$3,000) · Subscription (recurring): $200–$500 / lab / month · Enterprise (annual): $15K–$50K / year · Professional services: $5K–$20K per project.
Pitch Deck slide 14 — Business Model Canvas
Software-Only Pivot: moved the "Revenue Streams" and "Cost Structure" on Slide 13 away from hardware sales. The startup is now positioned as a SaaS that uses whatever materials are already in the lab.
MCV text-box "Summary of Changes Made" — direct response to committee comment

Why 7.5 / 10 — SolidSolid: 1,612% ROI is the most quantified financial case in the cohort + the SaaS pivot is exactly what the committee asked for. Held below 8.0 because the defensibility vs an incumbent (Thermo Fisher / ZEISS / Sartorius building this themselves) is asserted ("validated biological security through a software-first approach") rather than argued — what is the moat?

Presentation & Pitch /10i5.5 / 10

📐 ClarityMixed. Slide 4 architecture flow is clean. Slides 5–9 read as draft placeholder notes. Slide 14 BMC is wall-of-small-text. Format non-compliant: PPTX, not PDF. 14 content slides over the 10–12 cap.

📖 StorytellingNarrative arc is there — user persona ("Dr. Sarah") → emotional satisfaction curve → AI co-scientist framing → financial case → roadmap. But the storytelling is interrupted by the placeholder slides 5–9.

🎤 Persuasion$125K savings + 1,612% ROI is persuasive on its own. But no demo video means a reviewer can't see the system actually working. In-class pitch delivered.

Submitted as PPTX, not PDF (brief specifies PDF). 15 slides total / 14 content slides — over the 10–12 cap.
Compliance check vs brief p.3
Slide 14 Business Model Canvas — Key Partnerships, Key Activities, Key Resources, Customer Relationships, Channels, Customer Segments, Cost Structure, Revenue Streams all on one slide as small-text walls.
Pitch Deck slide 14 — density issue
No demo video provided — significant gap on a brief-listed deliverable that materially supports the Presentation criterion.
MCV submission inventory · 6871004530

Why 5.5 / 10 — AdequateLower-Adequate. Multiple compliance + clarity gaps stack: PPTX (not PDF) + 14 slides (over cap) + slides 5–9 placeholder + no demo video. The substantive content is OK; the execution polish is what holds the score back. A reviewer who only saw the deck (not the report) would underestimate the project.

AI Verdict · Claude Opus 4.7Hold — pending validation

This is the only project that fits the brief's hardware-integrated requirement, and the financial case is the most quantified in the cohort. But it has the worst validation gap of any submitted project — Slide 12 "Initial system feedback:" is literally empty, and no users are named anywhere. The Final Report cites itself as a primary reference, which is circular.

Fix first: interview 3 iPSC researchers (Chula / Mahidol / Siriraj have active labs), write a 1-page validation summary, and rebuild slides 5–9 from placeholder notes into proper investor slides. Integrity flag — moderate: the ROI math ($125K annual savings, 1,612% ROI) appears in the deck but no source data is shown; the AI-use acknowledgement is honest about heavy Gemini / Claude / Lovable usage, which is appropriate.

6 · Full Grading Report

The complete grading report contains: the full feedback adaptation tables (per group), the per-criterion scoring justifications with citations, the demo-video review (durations + content per video), and the cohort-level observations.

↗ Open the full grading report

Open Grading Spreadsheet (Google Sheets)