MDCU 3000747 — Healthcare Innovation & Entrepreneurship
Final Project Grading Report
Generated 22 May 2026 · Course weight under review: 40% (the rubric below totals 40%, matching the brief exactly)
Preamble — Conflicts & Methodology Notes
Grading is impartial and evidence-led. Every claim below cites the source file and section/slide. Where Feedback.docx is silent for a given student, that student is not penalised for absent source feedback (per the grader spec).
Conflicts between the grader system-prompt and the assignment brief — brief wins
- Total weight: System prompt asks me to "confirm with the brief whether the remaining 60% comes from other components". The brief (page 1 + grading rubric) explicitly states Weight: 40%, and the 5 rubric criteria sum to 40%. There is no "remaining 60%" — the rubric below is the entire 40%.
- Presentation & Pitch: System prompt says "focus on clarity; storytelling and persuasion secondary". The brief rubric explicitly weights "Clarity, storytelling, and persuasion skills" equally. I follow the brief.
- "Comment Response Document" as a required deliverable: The grader spec lists this as deliverable #6, but the assignment brief (page 3, Final Deliverables) does not list it. It exists as a committee-feedback follow-up. I check for it as a completeness item but I do not heavily penalise its absence.
- Hardware constraint: The brief states the solution "must be a physical product or hardware-integrated system (not just a mobile app)". Five of the six projects are app- or chatbot-only; only CellGuard AI describes a hardware-integrated system. Since the committee already reviewed all five at interim without rejecting them on this ground, I infer the constraint was relaxed in practice and do not penalise it in scoring. This is the single biggest discrepancy between the brief and what the cohort actually delivered.
Roster & group composition (derived from submissions + Feedback.docx)
| Student ID | Name | Project / Group | Status |
|---|
| 6678313530 | Xiaoyan Li | Project 3 — MindCheck (solo) | Submitted |
| 6778304030 | Chayanee Sae-lim | Project 6 — OonJai Breast Partners (solo) | Submitted; NOT in Feedback.docx |
| 6778310730 | Papawee Chennavasin | Project 5 — NutriTrack AI (paired with Kittitat) | Submitted (partial — addendum only) |
| 6871004530 | Zinab Omer | Project 4 — CellGuard AI (solo) | Submitted |
| 6878004830 | Khalilullah Arsyi | — | No submission |
| 6878005430 | Khin Sandar Win | — | No submission |
| 6878006030 | Khin Htet Htet | — | No submission (not even in MCV roster of submitters) |
| 6878301130 | Kittitat Jaidee | Project 5 — NutriTrack AI (paired with Papawee) | Submitted (team CEO) |
| 6878303430 | Chadaporn Attakitbancha | Project 2 — EV Optima AI (solo) | Submitted |
| 6878305730 | Xin Cai | Project 1 — Menopause Chatbot (solo) | Submitted |
Feedback Map (Step 2 — built from Feedback.docx)
Project 1 — Manopause Chatbot · Overall: Good
| # | Comment | Severity | Expected Revision |
|---|
| 1 | Revise User Journey with People–Process–Technology framework; emphasise human-in-the-loop; state long-term vision beyond Chatbot | Major | Add P-P-T diagram + HITL narrative + roadmap |
| 2 | Market scope too limited — extend beyond menopause; add e-commerce / telehealth / hospital integration | Critical | Reframe menopause as early-launch within broader women's health platform |
| 3 | Clarify personalisation features + data security; reference reliable medical info visually | Major | Add data fields used, retention/encryption details, source labels |
| 4 | Revenue capture if patient goes directly to hospital — how does the startup still make money? | Critical | Add referral/booking fee + SaaS B2B path |
| 5 | Missing pricing strategy (subscription/pay-per-use/sponsorship/referral) | Major | Concrete tiered pricing |
| 6 | Missing go-to-market plan | Major | Acquisition channels, partners, sequencing |
| 7 | APA references in main body | Minor | Add references slide / inline citations |
Project 2 — EV Optima AI · Overall: Excellent
| # | Comment | Severity | Expected Revision |
|---|
| 1 | Data acquisition strategy — sole reliance on own data insufficient; how/where will training data come from? | Critical | Multi-source data plan + collaborations |
| 2 | Spec-by-spec value-prop / competitor comparison; quantify economic benefits with vs without | Critical | Detailed feature table + cost/labour deltas |
| 3 | Map technology impact across the research → clinical → regulatory commercialisation pathway | Major | Pipeline-stage impact section |
| 4 | Missing pricing strategy | Major | Concrete tiers |
| 5 | Missing go-to-market | Major | Acquisition + channels |
| 6 | APA references in main body | Minor | Inline citation discipline |
Project 3 — MindCheck · Overall: Fair
| # | Comment | Severity | Expected Revision |
|---|
| 1 | Revenue capture if patient goes directly to hospital — clarify monetisation | Critical | Concrete revenue model |
| 2 | People–Process–Technology framework; human-in-the-loop; long-term vision beyond Chatbot | Major | P-P-T + HITL + roadmap |
| 3 | Opportunities beyond chatbot — limited market scope; add telehealth / hospital integration | Critical | Reposition chatbot as gateway to fuller platform |
| 4 | Spec-by-spec competitor analysis | Major | Detailed comparison table |
| 5 | Missing pricing strategy + clarify who pays (B2B universities?) + go-to-market (focus on hospital integration) | Critical | Concrete B2B+B2C pricing & channels |
| 6 | Slides too messy / too much text — review Visual Aids lecture | Minor | Visual cleanup |
| 7 | Verify AI-generated content before submission | Minor | Fact-check & edit |
| 8 | APA references properly cited | Minor | Reference hygiene |
Project 4 — Cell Guard · Overall: Good
| # | Comment | Severity | Expected Revision |
|---|
| 1 | Pinpoint the pain points (specific human errors e.g. oxygen drops, heater drying); calculate ROI on solving them | Critical | Explicit error list + ROI math |
| 2 | Spec-by-spec competitor benefits comparison | Major | Detailed comparison table |
| 3 | Clearly define the "predictive" future state | Major | Roadmap toward forecasting |
| 4 | Market positioning — concern it looks like an internal R&D project or feature for an existing equipment company; define unique value as a standalone startup | Critical | Defensible standalone-startup positioning |
| 5 | APA references properly cited | Minor | Reference hygiene |
Project 5 — NutriTrack · Overall: Excellent
| # | Comment | Severity | Expected Revision |
|---|
| 1 | Workflow integration — show exactly how the app fits into daily workflow | Major | Step-by-step workflow diagram |
| 2 | Caregiver-centric design — assume lower health literacy / education (especially Thai caregivers); keep intuitive | Major | Simplified UI + infographics |
| 3 | Spec-by-spec competitor analysis | Major | Detailed comparison table |
| 4 | Concrete numbers for TAM/SAM/SOM; focus on big cities first, not global; account for developed-country differences | Critical | Numbers + geographic prioritisation |
| 5 | Strategic partnerships — collaborate with dietary product / pharma / nutritionists | Major | Specific partnership categories |
| 6 | Missing pricing strategy + go-to-market | Major | Concrete tiers and channels |
| 7 | APA references properly cited | Minor | Reference hygiene |
Project 6 — OonJai Breast Partners (Chayanee)
No entry for "OonJai" or "Chayanee" exists in Feedback.docx. Per the grader spec, the feedback-adaptation check for this student is skipped, and absence of source feedback is not penalised. The student nonetheless self-describes a "Revised in response to faculty feedback" section addressing 5 areas (User Journey, Business Model, Revenue, Content Partnerships, Competitor Analysis), implying she received feedback through a different channel.
1. Executive Summaries (per group)
Group A — Menopause Chatbot (Xin Cai · 6878305730)
Project 1
Verdict: Solid submission. The committee asked for substantive rework of business model, scope, and personalisation; Xin Cai delivered on every point with traceable pointers (deck page + report section per comment) and visibly improved deliverables.
Top strengths: (1) Clean, well-designed pitch deck (17 pages incl. 4 backup) with consistent visual hierarchy and a working RAG chatbot at menochat.lovable.app; (2) genuine revision discipline — every committee comment is mapped to a specific deck page and report section in the Comments Responses document.
Top weaknesses: (1) Validation is minimal — 2 peer users, Likert scores only, no real menopausal user; (2) market scope still feels broad ("women aged 40–60, English-speaking, caregivers") without concrete TAM/SAM/SOM numbers.
Group B — EV Optima AI (Chadaporn Attakitbancha · 6878303430)
Project 2
Verdict: Strongest overall submission in the cohort. Already rated "Excellent" at interim; final deliverable maintains that bar and adds the missing data-acquisition strategy, spec-by-spec competitor table, and quantified ROI ($5,500 saving per cycle, 75% labour reduction).
Top strengths: (1) Tight problem-solution fit — a real B2B pain (EV optimisation inefficiency) with a working analytics MVP at ev-boost-insight.lovable.app; (2) the most quantitatively rigorous business case in the cohort — pricing tiers, 3-year revenue projection, before/after cost math.
Top weaknesses: (1) Validation is still only 2 EV researcher interviews (the brief's minimum); the platform has not yet been tested with real experimental data, which is acknowledged honestly; (2) deck runs 13 content slides — one over the 10–12 cap.
Group C — MindCheck (Xiaoyan Li · 6678313530)
Project 3
Verdict: Mixed. The pitch deck is conceptually thorough and addresses every committee comment in substance, but the submission package is incomplete: no separate Final Report file, no separate Comment Response Document, and validation is described as a future pilot plan rather than completed work.
Top strengths: (1) Strong problem grounding for Thai student mental health (10+ academic citations); (2) cleanest revenue-capture story among the chatbots — pay-per-referral ฿100–300 + ฿1,000–5,000 SaaS + university B2B subscriptions, with explicit answer to the "what if patient goes direct" question.
Top weaknesses: (1) Validation gap — slide 8 says "PILOT PLAN: at least >2 users" and slide 9 shows headers ("Results & Scoring") but no actual user data; (2) Final Report and Comment Response collapsed into the MyCourseVille text-box rather than separate files (committee asked for a written report).
Group D — CellGuard AI (Zinab Idris Omer · 6871004530)
Project 4
Verdict: The only project that genuinely fits the brief's "hardware-integrated system" requirement (cameras + edge capture + cloud AI + mobile alerts). Business case is quantified ($125K annual savings, 1,612% ROI), but key committee asks — competitor analysis depth, defensible standalone-startup positioning, and any actual validation — were addressed superficially.
Top strengths: (1) Genuine engineering-systems framing — the only deck that diagrams real hardware (slides 4–7); (2) financial validation is the most concrete in the cohort with explicit ROI math.
Top weaknesses: (1) Validation is essentially absent — slide 12 "Initial system feedback:" is empty, no named users, no validation summary file; (2) competitor analysis remains assertion-level ("unlike equipment giants…") with no spec-by-spec table the way EV Optima and NutriTrack delivered; (3) Comment Response is embedded as highlighted text in the report rather than a separate document.
Group E — NutriTrack AI (Kittitat Jaidee · 6878301130 + Papawee Chennavasin · 6778310730)
Project 5
Verdict: Excellent project, rigorously delivered. Already rated "Excellent" at interim; final deliverable adds a 7-step workflow integration diagram, caregiver-centric infographics (Thai/English toggle, how-to tabs), and a 7-feature spec table vs Tubie. The only material gap is validation thinness for a 2-person team.
Top strengths: (1) Highest-fidelity MVP in the cohort — full bilingual interactive flow at nutritrack-thai.lovable.app from patient input through titration plan to feeding diary; (2) detailed business model with concrete Thai-priced tiers (99–199 THB/month, 49 THB pay-per-7-day, B2B per-active-patient licensing) and partnership categories.
Top weaknesses: (1) Validation is 2 users (nurse + family caregiver) but the brief requires 2 per student → 4 for a team of 2; (2) TAM/SAM/SOM diagram still qualitative — slide 8 shows nested boxes with text labels but no concrete numbers despite committee asking specifically.
Group F — OonJai Breast Partners (Chayanee Sae-lim · 6778304030)
Not in Feedback.docx
Verdict: Conceptually strong and clinically credible (the founder is a breast surgeon MD), but the submission package is incomplete — no pitch deck in the latest version, only a revised report. The dual-purpose design (patient education + RWE registry) is novel.
Top strengths: (1) Strongest validation story in the cohort — 2 clinical reviewers with specific, before/after-implemented feedback (centralised menu, login/logout, family-history fields → BRCA-aware content); (2) phased business model (research phase via HSRI/KCMH grants → venture phase) is well-suited to the clinical-research context.
Top weaknesses: (1) No pitch deck in the final submission slot — only the revised DOCX report; this severely limits the "Presentation & Pitch" criterion; (2) No demo video link or public MVP URL in the submission.
2. Completeness Check (per student)
Key: ✅ found & compliant · ⚠️ found but non-compliant (see Notes) · ❌ missing. "Brief format" tracks the assignment-brief format expectations: PDF deck, ≥2 validation users/student, 5–7 min video, prototype, ≤1000-word report, separate comment-response document.
| Student ID | Name | Pitch Deck | Validation | Video | MVP | Report | Comment Response | File Location(s) | Notes |
|---|
| 6878305730 | Xin Cai |
✅ | ✅ | ✅ 5:49 | ✅ | ✅ | ✅ |
1_Menopause_Pitch_Deck_revised.pdf · 2_User_Validation_Result.pdf · 3_Menopause_Chatbot_Video.mp4 (5:49) · 4_MVP.txt → menochat.lovable.app · 5_…_Report_revised.pdf · 6_Comments_Responses_Document.pdf |
Deck: 11 content slides (within 10–12) + 4 backup (within ≤5). 2 peer users validated (minimum for solo). Comment response cross-references every committee point. Video duration 5:49 within 5–7 min spec. |
| 6878303430 | Chadaporn Attakitbancha |
⚠️ | ✅ | ⚠️ 4:01 | ✅ | ✅ | ✅ |
260423_Presentation_EVOpitimaAI_revised.pdf · EV_Optima_AI_Validation_Board_result.xlsx · Demo Google-Drive link (4:01) · ev-boost-insight.lovable.app · Final_Report.docx · Additional_Comment_Response_Document.docx |
Deck runs 13 content slides (one over the 10–12 cap). 2 EV researcher interviews documented (slide 9). Strong comment-response mapping to report sections. Video is 4:01 — under the 5-min minimum by ~1 min (content quality itself is strong). |
| 6678313530 | Xiaoyan Li |
⚠️ | ⚠️ | ⚠️ 4:27 | ✅ | ❌ | ⚠️ |
individual_presentation_xiaoyan_Li.pptx · Demo Google-Drive link (4:27) · mindlink-buddy.lovable.app · Comment response embedded in MCV text-box (5 numbered points) |
Deck submitted as PPTX not PDF (brief requires PDF). Deck slide 9 described validation as "PILOT PLAN" — but the video review shows the prototype is functional end-to-end (full PHQ-9 incl. Q9 suicidal ideation, GAD-7, two scored scenarios incl. emergency-red with crisis-support CTA, referral list). No separate Final Report file and no separate Comment Response Document. Video is 4:27 — under the 5-min minimum. |
| 6871004530 | Zinab Omer |
⚠️ | ❌ | ❌ | ✅ | ✅ | ⚠️ |
cell_guard_2.pptx · Final_Report.docx · Lovable URL in references · MCV text-box "Summary of Changes Made" (highlighted) |
Deck: PPTX (not PDF), 14 content slides (over 10–12 cap). No validation users named, no validation file. No demo video. Comment response embedded in text-box, not a standalone document. |
| 6878301130 | Kittitat Jaidee |
⚠️ | ⚠️ | ✅ 5:57 | ✅ | ⚠️ | ✅ |
01_NutriTrack_AI_Pitch_Deck.pdf · 02_Validation_Summary.pdf · Google-Drive demo link (5:57) · nutritrack-thai.lovable.app · 05_Final_report.pdf · 06_Comment_Response.pdf |
Deck: 14 content slides (over 10–12 cap). 2 validation users for a 2-student team (brief expects 4). Final report is 10 pages covering 5 prior assignments — likely well over 1000-word limit. Comment response is well-mapped to slide numbers. Video duration 5:57 within 5–7 min spec; content is cinema-grade (caregiver scenario + full prototype walkthrough + before/after impact). |
| 6778310730 | Papawee Chennavasin |
✅ (via team) | ⚠️ (via team) | ✅ (via team) | ✅ (via team) | ✅ (via team) | ✅ |
Additional_report.docx (her own contribution) + relies on team deliverables submitted under Kittitat's slot |
Her individual upload is only the Additional Report addressing committee feedback. Team's primary deliverables are filed under Kittitat. Acceptable team structure but reduces individual visibility. |
| 6778304030 | Chayanee Sae-lim |
❌ | ✅ | ⚠️ 3:15 | ⚠️ | ✅ | ⚠️ |
OonJai_Updated_Chayanee.docx (latest = revised report only) · Demo: YouTube youtube.com/watch?v=vjzJEt3xXGE (3:15) — supplied after initial review |
No pitch deck in latest submission (earlier "Slide_Project_Chayanee.pdf" from 22 Apr exists but was superseded — only the docx is in the latest slot). YouTube demo video exists but was outside the MCV submission slot; it's the shortest in the cohort (3:15, well under the 5-min floor). MVP described as "Lovable-built prototype" but no public URL given in the submission. Comment response is embedded in the revised report. OonJai is not in Feedback.docx — adaptation check skipped. |
| 6878004830 | Khalilullah Arsyi |
❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
— |
Appears in MCV submitter dropdown but submitted no attachments and no text-box content. |
| 6878005430 | Khin Sandar Win |
❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
— |
Same — empty submission slot. |
| 6878006030 | Khin Htet Htet |
❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
— |
Does not appear in either the MyCourseVille text-box listing or the attached-files listing. No submission of any form. |
2b. Demo Video Review
Videos were downloaded and inspected (6 frames per video sampled at evenly-spaced intervals + duration probed with ffprobe). The brief requires 5–7 minutes showing the problem, solution, and prototype in action.
| Project · Student | Duration | Spec (5–7) | Content observed (sampled frames) | Verdict |
|---|
| Menopause Chatbot · Xin Cai | 5:49 | ✅ |
Opens with "Current Gaps in Menopause Education" slide (4 pain blocks) → "Menopause Chatbot: An Intelligent Education Tool" overview → live MenoChat walkthrough: welcome screen with starter questions, full conversational answer covering GSM/HRT/MHT terminology, off-topic "What is Bitcoin?" deflection, and risky-question safety response to "I am so sad, I am depressed, I am afraid I may hurt..." with crisis hotlines. |
Problem → solution → prototype clearly demonstrated. Boundary control + safety responses visible. Consistent with the deck and report claims. |
| EV Optima AI · Chadaporn | 4:01 | ⚠️ |
Lab researcher scene with "Based on scientific data?" overlay → actual Excel cytokine dataset shown (Conditions × BDNF/EGF/FGF-2/HGF/IDO/IFN-γ/IL-4/IL-6/IL-10/LIF/PDGF-BB/SDF-1α/TNF/VEGF-A with highlighted cells) → live EVOptima AI dashboard with AI Recommendation panel and scores (33/36/32/35/0) → close-up of Optimization panel + Cytokine Weights sliders + Mitochondrial Support Insight → "How EVOptima AI work?" diagram → before/after framing (trial-and-error frustration vs data-driven dashboard). |
Content quality is strong but duration falls ~1 min short of the 5-min floor. Real experimental data shown, which is rare and impressive. |
| MindCheck · Xiaoyan Li | 4:27 | ⚠️ |
Hero illustration "Maybe... you don't need all the answers" with mindlink-buddy.lovable.app branding → side-by-side two scored scenarios: green "minimal to mild" 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 → walkthrough of actual PHQ-9 questions including Q9 (suicidal ideation: "Thoughts that you would be better off dead, or thoughts of hurting yourself") → GAD-7 questions (irritability) → final results screen PHQ-9=6/GAD-7=6 with referrals to Crisis Support / Find Mental Health Services / Bangkok Hospital / Siriraj Hospital. |
Duration short, but content is substantially stronger than the deck conveyed. The prototype demonstrably implements full PHQ-9 + GAD-7 + risk stratification + referral routing — score for MVP and Validation revised upward (see notes below). |
| NutriTrack AI · Kittitat + Papawee | 5:57 | ✅ |
Caregiver scenario shot — tired family member at bedside with notebook reading "FEEDING PLAN? How much formula? How often? Adjust when tolerance changes?" beside a tolerance log marked "bloating / nausea / better" → live nutritrack-thai.lovable.app walkthrough: Patient Information (weight/height/sex) → Formula Selection (Ensure / Peptamen / Blendera / Neomune with kcal/protein) → Nutrition Goal (30 kcal/kg/day, 1.2 g/kg/day) + Meal Schedule + Current Tolerance → Output (Total 2741 ml, Step-Up Titration Plan, Daily Water Plan +47 ml) → "Why NutriTrack AI Matters" infographic with three Before/After panels (Caregiver Confidence, Faster Nutrition Target Achievement 60%→100%, Fewer Unnecessary Hospital Visits) → brand close. |
Strongest demo video in the cohort. Full storytelling arc, working prototype, quantified impact framing, polished production. Within 5–7 min spec. |
| OonJai Breast Partners · Chayanee (YouTube link supplied after initial review) | 3:15 | ⚠️ |
Live walkthrough of a substantially developed Lovable web app branded "OonJai Breast Partners · Care, guided gently": (1) comprehensive oncology intake — Clinical stage (TNM with T1/T2/T3/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); (2) Family & genetic history with first-degree-relative selector + BRCA1/2 status + PDPA consent banner + "Create my private space" CTA; (3) "Stage II — In active treatment" dashboard with Ask Guide → "OonJai Guide" answering from NCCN Breast Cancer Guidelines v.2.2026 with 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?"); (4) AI guide rendering a detailed HR+/HER2- explanation with section headers and citation; (5) Monthly Check-in PRO questionnaire (nausea / fatigue / skin or hair changes / numbness or tingling, 5-point Likert); (6) Follow-up status form with patient-state tracking (NED · Local recurrence · Regional recurrence · Distant metastasis). |
Duration well below the 5-min floor — the shortest video submitted. But the MVP is by some distance the most clinically sophisticated in the cohort: NCCN-grounded oncology education + structured intake + PRO collection + recurrence tracking is a real clinical-research infrastructure prototype, not a demo. MVP score revised up after seeing this video (see notes below). |
Score revisions following video review:
- MindCheck (Xiaoyan): MVP revised 3.5 → 4.0 — video shows end-to-end working PHQ-9 + GAD-7 with emergency-red flow and Bangkok/Siriraj referrals. Validation stays at 2.0 (developer demo, not user testing). New total 25.5 → 26.0.
- OonJai (Chayanee): MVP revised 3.5 → 4.5 — video shows a clinically credible NCCN-grounded staging intake + AI education guide + PRO check-in + recurrence-status logging, exceeding what the report alone conveyed. New total 27.0 → 28.0.
3. Feedback Adaptation Tables (per group)
Group A — Menopause Chatbot
| Committee Comment | Severity | Acknowledged? | Revised? | Location Cited? | Quality |
|---|
| P-P-T framework + HITL + long-term vision | Major | Y | Y | Y (Deck p.6 + p.9; Report §2) | Substantive — explicit "Human-in-the-Loop Care Journey" diagram (slide 6) with People/Process/Technology roles mapped |
| Beyond Menopause / expand market | Critical | Y | Y | Y (Deck p.9; Report §2) | Substantive — Slide 9 "From Menopause Chatbot to Women's Health Platform" with 4-stage scope expansion |
| Personalisation features + data security | Major | Y | Y | Y (Deck p.5; Report §3) | Partial — Slide 5 lists "Privacy Protection" + "Data Security" cells; Report §3 mentions consent, encryption, de-identified analytics, but specific retained-data fields list is shallow |
| Revenue capture if patient goes direct to hospital | Critical | Y | Y | Y (Deck p.8; Report §6) | Substantive — Slide 8 stream 4: "Referral and Booking Partnership… Capture value even when care is delivered by external hospitals or clinics" |
| Pricing strategy | Major | Y | Y | Y (Deck p.8; Report §6) | Substantive — explicit 100 THB/mo, 250 THB/qtr, 800 THB/yr tiers + B2B SaaS + integration fee + per-success referral |
| Go-to-market | Major | Y | Y | Y (Deck p.9; Report §7) | Substantive — 3-phase market entry: international hospitals → community outreach → digital campaigns |
| APA references in body | Minor | Y | Y | Y (Deck p.12 References) | Substantive — references slide added; in-text [n] citations consistent |
Group B — EV Optima AI
| Committee Comment | Severity | Acknowledged? | Revised? | Location Cited? | Quality |
|---|
| Data acquisition strategy | Critical | Y | Y | Y (Report §3.1–3.3; Deck p.5) | Substantive — 4 data sources (internal/academic/industry/public DBs) + value-exchange collaboration model + continuous learning loop |
| Spec-by-spec competitor comparison + quantified benefits | Critical | Y | Y | Y (Report §7; Deck p.10) | Substantive — 6-row comparison vs Benchling/Dotmatics/JMP/traditional; iteration time, personnel, conditions all quantified |
| Economic benefits with vs without | Critical | Y | Y | Y (Deck p.11) | Substantive — explicit baseline $8,000/cycle vs $2,500/cycle, $5,500 net saving, 75% labour / 50% time / 70% cost reduction |
| Research commercialisation pathway impact | Major | Y | Y | Y (Report §8) | Substantive — Discovery → Preclinical → Clinical Translation → Clinical Trials → Regulatory mapped explicitly |
| Pricing strategy | Major | Y | Y | Y (Report §5.1; Deck p.8/13) | Substantive — hybrid SaaS ($30/mo + $10/analysis + enterprise custom), ~$80/lab/mo average |
| Go-to-market | Major | Y | Y | Y (Report §6; Deck p.12) | Substantive — conferences (ISEV, ISCT), academic collaborations, B2B engagement; 5-stage funnel |
| APA references | Minor | Implicit | Partial | — | Partial — references present at end of report but in-text [n] citation style is uneven |
Group C — MindCheck
| Committee Comment | Severity | Acknowledged? | Revised? | Location Cited? | Quality |
|---|
| Revenue capture if patient goes direct | Critical | Y | Y | Y (text-box pt.1 → Deck p.7) | Substantive — pay-per-referral ฿100–300 + SaaS ฿1,000–5,000/clinic + university plans |
| P-P-T + HITL + vision beyond chatbot | Major | Y | Y | Y (text-box pt.2 → Deck p.5, p.7) | Substantive — Slide 7 names "Human-in-the-Loop" + AI risk stratification + smart referral; Slide 5 "Beyond Chatbot: Scalable Healthcare Platform Vision" |
| Opportunities beyond chatbot (telehealth, hospital integration) | Critical | Y | Y | Y (text-box pt.3 → Deck p.5) | Substantive — Slide 5 explicitly adds Telehealth Integration, Hospital/Clinic Integration, Mental health marketplace |
| Spec-by-spec competitor analysis | Major | Y | Y | Y (text-box pt.4 → Deck p.10) | Substantive — explicit "OOCA solves 'I want therapy' / We solve 'Do I need therapy?'" + Assessment→Score→Risk→Action path |
| Missing business fundamentals (pricing, B2B vs B2C, GTM) | Critical | Y | Y | Y (text-box pt.5 → Deck p.7, p.11) | Substantive — pricing tiers explicit; B2B clinics primary payer, B2B universities secondary, students as users not primary payers; 3-channel GTM |
| Slides too messy / too much text | Minor | N (implicit) | Partial | — | Partial — slides 6 and 7 remain dense with multiple sub-blocks; visual cleanup uneven |
| Verify AI-generated content | Minor | N | Unclear | — | Ignored — no statement of AI verification process |
| APA references | Minor | Implicit | Y | Per slide | Substantive — references appear per slide with numbered citations |
Group D — CellGuard AI
| Committee Comment | Severity | Acknowledged? | Revised? | Location Cited? | Quality |
|---|
| Pinpoint specific human-error pain points + ROI | Critical | Y | Y | Y (text-box "Technical Detail" → Slide 3; "Financial Proof" → Slide 8) | Substantive — oxygen drops, heaters drying, overnight growth speed all named; $125K annual savings + 1,612% ROI; $2,500–$5,000 per failed batch |
| Spec-by-spec competitor comparison | Major | Y | Partial | Y (text-box "Competitive Edge" → Slide 10) | Partial — Slide 10 is value-prop list, not a spec table; report claims "SaaS vs equipment giants" but provides no feature-by-feature comparison the way EV Optima or NutriTrack did |
| Define "predictive" future state | Major | Y | Y | Y (text-box "Future Vision" → Slide 14) | Substantive — Phase 3 "Virtual Monitoring" with software-only time-based simulations forecast in advance |
| Market positioning vs internal R&D project / equipment-company feature | Critical | Y | Partial | Y (text-box "Software-Only Pivot" → Slide 13) | Partial — pivots to "SaaS that uses whatever materials are already in the lab" (no new hardware), which is the right move; but the defensibility argument vs an incumbent (Thermo, ZEISS, etc.) building this feature is asserted, not argued |
| APA references | Minor | Y | Y | Report references list | Substantive — APA-style references list included; in-text citation style is unconventional (e.g. "Idris, Z. (2026)." with trailing period) but present |
Group E — NutriTrack AI
| Committee Comment | Severity | Acknowledged? | Revised? | Location Cited? | Quality |
|---|
| Workflow integration in daily workflow | Major | Y | Y | Y (Deck p.15) | Substantive — 7-step diagram: Hospital assessment → Discharge onboarding → Caregiver input → AI feeding plan → Home feeding → Daily tracking → Clinical review |
| Caregiver-centric design for lower-literacy users | Major | Y | Y | Y (Deck p.12–14; Papawee Addendum) | Substantive — "How to use this app" infographic tabs (5 steps with icons), Thai/English toggle, visual formula cards with kcal/protein per scoop |
| Spec-by-spec competitor comparison | Major | Y | Y | Y (Deck p.11; Comment Resp.) | Substantive — full NutriTrack vs Tubie table across 7 dimensions with explicit "Benefit with NutriTrack" column |
| Concrete TAM/SAM/SOM numbers + geographic focus | Critical | Y | Partial | Y (Deck p.8; Report p.3) | Partial — narrows geography to Bangkok→Thailand (good), but the slide is still a nested-box diagram with text labels rather than dollar/patient counts; report mentions global $7–8B + Thai $62.4M but doesn't compute SAM/SOM headcount |
| Strategic partnerships with dietary / pharma / nutritionists | Major | Y | Y | Y (Deck p.10; Papawee Addendum) | Substantive — explicit Nestlé/Abbott partnership concept with one-click reorder, brand-sponsored adherence reports, 3–5% referral commission |
| Pricing strategy + go-to-market | Major | Y | Y | Y (Deck p.9; Report p.10) | Substantive — 14-day free trial + 99–199 THB/mo + 49 THB/7-day pay-per-use + B2B hospital licensing per active patient + brand-neutral DB sponsorship; KCMH pilot → Bangkok → home-care networks → LINE OA |
| APA references | Minor | Y | Y | Report Reference § | Substantive — references list with DOIs; in-text [n] style used |
Group F — OonJai Breast Partners
Adaptation check skipped — no entry exists in Feedback.docx for Chayanee Sae-lim's OonJai project. The student self-describes a 5-area revision (User Journey, Phased Business Model, Diversified Revenue, Content Partnership, Competitor Matrix) and demonstrates clear before/after evidence (Figure 2 vs 3 navigation menu, family-history field addition), suggesting feedback was received but not captured in the document I was given.
4. Detailed Rubric Scoring (per group, group-level)
Group A — Menopause Chatbot (Xin Cai)
Criterion (Weight)
Score
Justification (cited)
Improvement Suggestion
Problem-Solution Fit (10)
7.5
Real, well-cited problem (5_…_Report §1 cites Lancet 2024, IMS 2025, McCartney 2022 BMJ); solution maps to it (RAG chatbot grounded in IMS guidelines, with boundary control + safety responses for risk situations, Report §2–3). Customer insight is thin — only 2 peer users, neither an actual menopausal woman.
Recruit ≥3 women in perimenopause (45–55) for the next validation round; have them rate factual accuracy of chatbot answers, not just usability.
MVP (5)
4.0
Working public Lovable prototype at menochat.lovable.app; Deck p.15–16 shows login flow, welcome screen with starter questions, non-related-question handling, and risky-question safety response with 3 emergency hotlines. RAG architecture described (Report §3).
Add evidence the knowledge base is actually built from IMS/NAMS guidelines (currently asserted but not demonstrated — show indexed source list).
Validation Evidence (5)
3.5
2 female peer users with Likert scale (Usability 5/5, Clarity 4/4, Willingness 4/3) and qualitative open-ended feedback (Deck p.17). User 2's "scope should be broader" feedback is visibly the driver of the platform-vision pivot in Deck p.9 — clear "we changed X because user said Y" pattern.
Validation users were peers, not target demographic — invalidates external validity. Run a second round with 5+ women in menopause age range before any pilot.
Business Viability (10)
7.5
4 distinct revenue streams with Thai pricing (Deck p.8): B2C freemium 100 THB/mo; B2B SaaS for clinics/hospitals; telehealth integration fee + commission; pay-per-success referral. Direct answer to "what if patient goes direct" via stream #4 (Report §6).
No TAM/SAM/SOM numbers — slide 4 says "~1 billion women globally by 2030" but no Thai market sizing. Add concrete Thai perimenopausal population estimate and conversion assumptions.
Presentation & Pitch (10)
8.0
Consistent peach/teal/yellow palette; clean iconography; one-idea-per-slide layout; structured narrative (problem → market → solution → P-P-T → business → revenue → vision → competition → impact → references + 4 backup); within slide-count cap.
Slide 11 "Impact" reads as feel-good marketing without measurable impact metrics — replace soft phrases ("greater confidence") with target numbers (X% reduction in time-to-care, Y% increase in literacy score).
Subtotal: 30.5 / 40 · 76.25%
Group B — EV Optima AI (Chadaporn)
Criterion (Weight)
Score
Justification (cited)
Improvement Suggestion
Problem-Solution Fit (10)
8.5
B2B EV-production inefficiency is a real, expensive R&D pain (Deck p.2–3); solution (predictive AI ranking experimental conditions) directly attacks trial-and-error waste. Problem framing validated through EV researcher interviews (Deck p.9; Report §1). Three core pains explicitly enumerated.
The "B2B SaaS for academic labs" framing is correct but academic labs are not strong payers — interview a CDMO/biotech procurement lead to test whether the $30/mo + $10/analysis price actually clears their procurement bar.
MVP (5)
4.5
Highest-fidelity analytics MVP in the cohort (ev-boost-insight.lovable.app, Deck p.6): radar plots, cytokine heatmap, AI-assisted condition recommendation with key drivers, EV score ranking, custom weighting sliders. Validation Board xlsx contains real cytokine experimental data.
The AI today is rule-based scoring on uploaded data, not a trained predictive model — be transparent in the pitch that this is a decision-support skin until enough training data is collected.
Validation Evidence (5)
4.0
2 EV researchers interviewed (Deck p.9, Report §1); pain points triangulated ("too many experiments, not enough direction", "high cost, low reproducibility", "complex data interpretation"). Continuous Learning Loop (User Input → AI Prediction → Validation → Re-upload → Refinement) wired in (Report §3.3) — exactly the "iterate based on feedback" pattern the rubric rewards.
Validation is interview-only; no user has yet used the platform on their own data. Recruit 1 academic lab for a 4-week pilot with real cytokine datasets before pricing.
Business Viability (10)
8.5
Most quantitatively rigorous in the cohort: hybrid pricing (Report §5.1, Deck p.8/13), 3-year revenue projection ($10K/$38K/$115K with 10→40→120 labs, Report Table), explicit baseline-vs-with cost math ($8,000 → $2,500 per cycle = $5,500 net saving; Deck p.11). Multi-source data acquisition + value-exchange partnership model addresses the Critical interim ask.
120 labs by Year 3 is aggressive given ISEV community size — add a sensitivity case (50% adoption) and articulate which 10 Y1 labs are realistically winnable.
Presentation & Pitch (10)
8.0
Professional design, strong narrative arc (EV explained → core challenge → solution → market → business → traction → competition → marketing → financial → vision). Deck p.2 is a model of dense-information-done-right with three-column "What/Why/Challenge" framing.
13 content slides — one over the 10–12 cap. Trim the duplicate "Solution" slides (p.4 + p.5 + p.6) into one. Slide p.3 has unreadable raw data tables — convert to one summary number.
Subtotal: 33.5 / 40 · 83.75%
Group C — MindCheck (Xiaoyan Li)
Criterion (Weight)
Score
Justification (cited)
Improvement Suggestion
Problem-Solution Fit (10)
6.5
Problem is real and densely cited (Deck p.3: depression 9.3–30.5% among medical students, stress 61.4%, suicidal ideation 12.8%, plus cultural stigma + low awareness, 10 academic refs). Solution (PHQ-9/GAD-7 screening → AI risk stratification → referral) follows clinical logic.
"Customer insight" is largely literature review, not primary interviews. Before pilot, recruit 3–5 Thai university students and 1–2 clinic operators to test the actual flow.
MVP (5)
4.0
Web prototype at mindlink-buddy.lovable.app (Deck p.8). Decision pathway diagram (Deck p.10) shows Assessment → Score → Risk → Action mapping with mild/moderate/severe stratification. Video review confirms a substantially more complete prototype than the deck conveys: full PHQ-9 questionnaire (including Q9 suicidal ideation), full GAD-7, two scored end-states (mild-symptom green + emergency-red with Crisis Support CTA), and a tiered referral list (Crisis Support · Find Mental Health Services · Bangkok Hospital · Siriraj Hospital). Revised upward from 3.5 after video inspection.
Add screenshots from the video into the deck (Deck p.8 currently shows none) — the working prototype is the team's biggest underclaimed asset.
Validation Evidence (5)
2.0
Major weakness — score unchanged after video review. Deck p.8 frames validation as future ("PILOT PLAN: 1 University pilot site, 100–300 users internal testing (at least >2)"); p.9 lists method headers but no actual user data. The video is a developer-led system demo, not user testing with real Thai students — no user identities, no user reactions, no Likert data, no "we changed X because user said Y" pattern. Brief requires 2 users actually tested.
Run the validation now — even 2 user-tests with PHQ-9/GAD-7 walkthrough on the live Lovable prototype, and capture the "we changed X because user said Y" pattern explicitly.
Business Viability (10)
7.5
Best-articulated revenue capture among the chatbot teams (Deck p.7): pay-per-referral ฿100–300, SaaS ฿1,000–5,000/clinic/mo, university B2B plans, optional student premium. Explicit answer to the Critical "what if patient goes direct" comment (text-box pt.1) — capture via SaaS even without referral. 3-channel GTM (hospital integration, university channel, digital).
The "Universities pay" assumption needs validation — Thai universities have historically resisted paying for mental-health vendor tools. Add at least one university stakeholder interview.
Presentation & Pitch (10)
6.0
Slide count compliant (11 content excluding cover). However, the committee flagged "slides too messy" at interim and several slides remain dense: Slide 6 has 5 stacked statistics blocks; Slide 7 packs the whole business model + revenue capture logic + pricing overview onto one slide; Slide 10 attempts to layer self-help/therapy/triage narratives.
Apply the "one idea per slide" principle to slides 6, 7 and 10; split slide 7 into "Business Model" + "Pricing" + "Customer Segments" as three slides.
Subtotal: 26.0 / 40 · 65.0% (revised up from 25.5 after video review increased MVP score from 3.5 to 4.0)
Group D — CellGuard AI (Zinab)
Criterion (Weight)
Score
Justification (cited)
Improvement Suggestion
Problem-Solution Fit (10)
7.0
Real lab pain (iPSC overnight blind spot 6 PM–8 AM with $2,500–$5,000 per failed batch, Report §2–3; Deck p.3 "Critical Intervention Zones"). The only project in the cohort that genuinely fits the brief's "hardware-integrated system" requirement — cameras + edge capture + cloud AI + mobile alerts (Report §3; Deck p.4 "AI-Assisted Monitoring" flow). User persona "Dr. Sarah" is one composite, not multiple validated stakeholders.
Talk to 3–5 actual iPSC researchers (Chula Stem Cell + Mahidol + Siriraj have active labs) and capture variance in their actual monitoring practice before claiming "40% of working hours" is industry-wide.
MVP (5)
3.5
Lovable dashboard URL exists (Report references). Slide 6 shows tabs/factors/visual-input architecture; Slides 7–10 show evaluation flow concept.
The slide text is fragmentary ("Connect the visual input device to the Ipscs (system)" on Slide 5) — these read like placeholder notes, not investor-ready slides. Rebuild Slides 5–9 with annotated screenshots.
Validation Evidence (5)
1.5
Severe weakness. Slide 12 "Initial system feedback:" is empty. No validation summary file submitted. Final Report §2 refers to "validation research… to quantify market need" but cites only market-report figures, not user testing. No named interviewees, no Likert data, no "we changed X because user said Y" pattern.
This is the single biggest grade-pull on the submission. Even 2 interviews with iPSC researchers — captured in a 1-page summary — would unblock most of the lost score.
Business Viability (10)
7.5
Quantified financials: $125K annual savings, 1,612% Y1 ROI, $7,300 annual cost (Slide 8). BMC complete on Slide 14: $200–$500/lab/mo subscription, $15K–$50K enterprise, $5K–$20K professional services, $1.5K–$3K implementation fee. Software-only pivot per committee feedback is the right move (Slide 13 references).
The defensibility argument vs an incumbent (Thermo Fisher, ZEISS, Sartorius) building this feature themselves is asserted, not argued — what is the moat? Curated iPSC morphology dataset? Then quantify how many labelled images you have and aim to acquire.
Presentation & Pitch (10)
5.5
Deck submitted as PPTX, not PDF. 14 content slides excluding cover (over 10–12 cap). BMC slide is a wall of small text. Slides 5–9 read as draft placeholder notes ("Tabs for navigation", "Type of systems", "Visual input", "factors input" with no body content extracted). Inconsistent layout density across slides.
Cut the deck to 10–12 slides by collapsing 5–9 into one "How It Works" slide with a single diagram + 3 callouts. Convert to PDF.
Subtotal: 25.0 / 40 · 62.5%
Group E — NutriTrack AI (Kittitat + Papawee)
Criterion (Weight)
Score
Justification (cited)
Improvement Suggestion
Problem-Solution Fit (10)
8.5
Clinically credible problem (Discharged-on-Enteral-Nutrition readmission risk, top causes documented in Current Gastroenterology Reports 2023, Deck p.3). Founded by physicians who observed the gap themselves (Report Assignment 1). Specific use cases (post-operative, oncology). The "AI does what only ICU hardware or hospital doctors currently do, but at home" framing on Slide 15 is sharp.
Calculate the actual harm number — how many readmissions per 1,000 enteral patients per year in Thailand, and what fraction could plausibly be averted with titration support? Use that to dimension the saved-DALY/avoided-readmission cost.
MVP (5)
4.5
Highest-fidelity MVP in the cohort (nutritrack-thai.lovable.app). Slides 12–14 walk through patient input → formula selection (Ensure/Peptamen/Blendera/Neomune with kcal/protein) → goal + fluid loss → tolerance entry → titration plan → 3-meal-prep guidance with water flush → daily diary tracking with %target compliance. Thai/English language toggle.
The titration logic depends on a medical food database that doesn't exist yet (Report §1, identified as the major challenge). Quantify how many SKUs are required for Bangkok hospital coverage and what data-licensing path is realistic.
Validation Evidence (5)
3.0
2 validation users documented in 02_Validation_Summary.pdf: hospital nurse + family caregiver. Both gave structured feedback that visibly drove revisions — discharge feeding summary, safety alerts, step-by-step prep guide, reminders, daily intake log all appear in the revised UX. However, brief requires 2 per student → 4 total for a 2-person team. Only half delivered.
Recruit 2 more validators before showcase — ideally a clinical dietitian (KOL) and a second caregiver type (e.g. oncology patient family) — to hit the brief's explicit requirement.
Business Viability (10)
8.5
Most concrete monetisation in the cohort (Deck p.9): 14-day trial → 99–199 THB/mo subscription, 49 THB pay-per-7-day plan, B2B hospital licensing per active patient, brand-neutral medical-food-DB sponsorship, 3–5% referral commission on in-app reorders. Tubie competitor analysis (Deck p.11) is the cohort's gold standard with 7 dimensions × 3 columns (NutriTrack / Tubie / Benefit). KCMH pilot → Bangkok expansion → LINE OA + distributor channels.
TAM/SAM/SOM is qualitative ("Thailand market / Bangkok hospitals / KCMH") despite committee explicitly asking for concrete numbers. Add patient counts and TH-Baht market size to the SOM and SAM rings.
Presentation & Pitch (10)
8.0
Consistent green palette, clean infographics (7-step workflow integration on Slide 15 is the cohort's best single graphic), Tubie comparison table (Slide 11) is clear and persuasive, caregiver-centric "How to use" tabs (Slide 12) and screen-shot grid (Slide 13–14) make the product tangible.
14 content slides (over 10–12 cap). Slide 6 "Key Features" with only 2 cards (Smart Titration + Real-time Adjusts) is a stub; either expand to 4 features or merge into Slide 5.
Subtotal: 32.5 / 40 · 81.25%
Group F — OonJai Breast Partners (Chayanee)
Criterion (Weight)
Score
Justification (cited)
Improvement Suggestion
Problem-Solution Fit (10)
7.5
Clinically credible — founder is a breast surgeon MD addressing real treatment-pathway confusion. Dual-purpose design (patient education + structured RWE/QOL/survival registry via LINE OA) is genuinely novel and aligned with Thai oncology research infrastructure. Report §2 maps patient journey across 5 clinical stages with stage-by-stage success metrics.
The dual-purpose pitch is also a dual-purpose risk — patients may not engage if they sense data extraction. Strengthen the consent + IRB narrative.
MVP (5)
4.5
Revised up after video review. The YouTube demo (3:15) shows a clinically sophisticated Lovable web app: comprehensive oncology intake (TNM staging, ER/PR/HER2 subtype, surgery type, reconstruction options, axillary surgery, family/genetic history with BRCA status, PDPA consent), an AI guide grounded in NCCN Breast Cancer Guidelines v.2.2026 with sophisticated suggested questions (HR+/HER2-, SLNB vs ALND, CDK4/6 inhibitors), monthly PRO check-in (nausea / fatigue / skin or hair / numbness), and follow-up status logging with recurrence categories. This is closer to a real clinical-research infrastructure than a class prototype.
Surface the working URL into the submission slot — the deliverable underclaims what was built. A two-line README pointing to the demo video + Lovable URL would close the gap.
Validation Evidence (5)
4.0
Strongest validation narrative in the cohort despite small N: 2 clinical reviewers with specific actionable feedback explicitly implemented and visualised with before/after (Figure 2 vs 3 navigation, Figure 4–5 family-history field driving BRCA-aware content). Crystal-clear "we changed X because user Y said Z" pattern (Report §9).
Reviewer 1 + Reviewer 2 are clinicians, not patients. Run a second round with 2–3 actual breast cancer patients (KCMH oncology clinic) to validate the education content itself.
Business Viability (10)
8.0
Phased model is the smartest fit for a clinical-research-driven product (Report §3): Phase 1 research via HSRI grant + KCMH research budget; Phase 2 venture via B2B/B2C revenue. Diversified revenue (Report §4 — freemium + sponsorship + B2B partnerships) + governance structure (independent medical board, transparent content labelling) addresses pharma-bias risk credibly. Competitive matrix vs Outcomes4Me/PatientsLikeMe/Belong.Life (Report §6) is well-scoped.
Thai TAM is small (~17K new breast cancer cases/year per NCCN 2023) — model scalability via SE-Asia regional or via expansion to other cancers (mentioned but not quantified).
Presentation & Pitch (10)
4.0
No pitch deck in the latest submission. The earlier 22-Apr "Slide_Project_Chayanee.pdf" was superseded by the docx report and is not in the latest slot. Without a deck, the criterion (which is specifically about a 10–12 slide investor pitch) cannot be scored above "Weak". The revised report itself is well-organised, but it is not a pitch.
The simplest fix: re-export the existing slides as PDF and re-submit. Without this, the project loses ~5 points in this criterion despite strong underlying substance.
Subtotal: 28.0 / 40 · 70.0% (revised up from 27.0 after video review increased MVP score from 3.5 to 4.5)
5. Total Scores (group level)
| Group | Project | 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 (revised after video review) | 7.5 | 4.5 | 4.0 | 8.0 | 4.0 | 28.0 | 70.0% | Adequate → Solid |
| C | MindCheck · Xiaoyan Li (revised after video review) | 6.5 | 4.0 | 2.0 | 7.5 | 6.0 | 26.0 | 65.0% | Adequate |
| D | CellGuard AI · Zinab (no video provided) | 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): no group, no submission, 0 / 40 on the group-project component. |
Banding (per system-prompt scale): 90–100% Exceptional · 75–89% Solid · 60–74% Adequate · 40–59% Weak · 0–39% Inadequate. No group reaches the Exceptional band; two cluster in upper-Solid, one mid-Solid, three in the Adequate range.
6. Grader's Opinions (honest verdict, per group)
Group A — Menopause Chatbot (Xin Cai)
Fundable: conditional
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.
Group B — EV Optima AI (Chadaporn)
Fundable: most defensible in cohort
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.
Group C — MindCheck (Xiaoyan Li)
Fundable: not yet
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.
Group D — CellGuard AI (Zinab)
Fundable: hold 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 ("Idris, Z. (2026). CellGuard AI: Precision automation… [Unpublished manuscript]") 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.
Group E — NutriTrack AI (Kittitat + Papawee)
Fundable: closest to a real product
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.
Group F — OonJai Breast Partners (Chayanee)
Fundable: depends on missing deck
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 navigation + BRCA family-history field) is the cleanest "we changed X because user Y said Z" evidence in the cohort.
Cohort-level observations
- Hardware constraint largely ignored. 5 of 6 projects are app- or chatbot-only despite the brief stating "must be a physical product or hardware-integrated system, not just a mobile app". The committee accepted these at interim, so the constraint was implicitly relaxed. Worth flagging in next semester's brief if this is intentional.
- Lovable.app is the de-facto cohort MVP tool. All 6 working prototypes are Lovable web apps. Worth either embracing this (and updating the brief to recognise no-code prototyping as valid) or pushing teams toward more diverse stacks.
- Validation discipline is the most uneven dimension. Scores ranged from 1.5/5 (CellGuard, essentially absent) to 4.0/5 (OonJai, EV Optima). Three teams met the bare minimum (2 users); CellGuard didn't, and MindCheck described validation as a future plan.
- Three students submitted nothing. Khalilullah Arsyi, Khin Sandar Win, and Khin Htet Htet have no deliverables. If they are still enrolled, they should receive 0 / 40 on this component unless special circumstances apply.
- Comment Response Document is a deliverable not listed in the assignment brief. 3 teams provided a separate file (Xin Cai, Chadaporn, Kittitat); 3 embedded it in the text-box / report (Xiaoyan, Zinab, Chayanee). If this is meant to be required, add it to the brief.
- Demo video compliance is mixed. 5 of 6 teams provided videos (Xin Cai 5:49 ✅ · NutriTrack 5:57 ✅ · MindCheck 4:27 ⚠️ · EV Optima 4:01 ⚠️ · OonJai 3:15 ⚠️, supplied post-submission via YouTube link). Only CellGuard has no video at all. Three videos fall below the 5-min floor — consider whether to enforce strictly or treat as soft. Several MVPs are materially more developed than the static deliverables convey — particularly MindCheck (PHQ-9/GAD-7 + emergency-red flow) and OonJai (NCCN-grounded oncology platform). If videos are not reviewed, those teams are under-credited.
End of report · Generated by Claude (Opus 4.7) acting as impartial grader · All claims cited from primary deliverables in D:\MED Final Submission\submissions\