Problem Statement
After every habagat, thousands wade in floodwater; a week later fevers spike and RHUs overflow. Families don’t know leptospirosis signs (lagnat + sakit ng binti + paninilaw) and wait until dialysis is needed. RHU staff repeat the same 5 questions by phone while severe cases wait.
Proposed Solution
Answer 5 taps: lumusong ba, may sugat, ilang araw ng lagnat, sakit ng binti, ihi/kulay. LLM triage gives color-coded advice (“PULA — pumunta sa RHU ngayon, dalhin ang listahang ito”) plus nearest RHU hours and transport tip. Severe cases jump the RHU queue; mild get home-care checklist and 48h callback reminder via SMS.
Target Market
- Primary users: Flood-waders, parents, seniors in NCR, Bulacan, Pampanga.
- B2G: RHUs, barangay health workers, DOH triage lines.
How AI Is Used
- Symptom reasoner: LLM maps exposure + symptoms to DOH-aligned urgency; rules alone miss Taglish phrasing.
- Dialect handling: Understands Bisaya/Tagalog fever descriptions.
PLDT / Smart / ePLDT / Rokid Angle
- Smart: SMS triage for no-data users; voice readout for seniors.
- ePLDT: Hosts RHU queue dashboard in-country.
Demo Scenario (3 minutes for judges)
- Mock case: waded 3 days ago, fever + leg pain; verdict PULA with RHU slip.
- Mild case: no wound, 1-day low fever; verdict DILAW with home care.
- SMS version returns same advice on basic phone.
- RHU queue shows severe first.
- Close: “Bago lumala, mag-triage.”
30-Hour Build Plan
| Hours | Task | Role |
|---|---|---|
| 0–3 | DOH leptospirosis guides + RHU mock directory | All |
| 3–12 | Triage prompts + safety rails | AI lead |
| 3–10 | Chat + SMS + queue | Backend |
| 10–20 | Family-first UI | Frontend |
| 20–26 | Clinician review of outputs | All |
| 26–30 | Pitch, disclosure | PM / Pitch |
Suggested stack: Chat UI + LLM triage, RHU directory, SMS fallback, Postgres queue. Public data / APIs: DOH leptospirosis advisories (public guidance).
Risks & Privacy Notes
- Medical safety: Triage only, never diagnosis; always “see health worker,” low-confidence defaults to PULA. Demo cases synthetic; no real health data into AI tools.
- Privacy: Health answers minimized, auto-purged in 30 days, consent-first.
AI Use Disclosure (draft)
Uses LLM triage reviewed against DOH guidance. Coding assistants helped with chat UI. All demo outputs clinician-style checked by the team.
Rubric Scoring
| Criterion | Weight | Score (1–5) | Why |
|---|---|---|---|
| Innovation & creativity | 25% | 3 | Triage bots exist; flood-specific + SMS matters |
| Problem–solution fit | 25% | 5 | Post-baha deaths preventable |
| AI application & value | 20% | 4 | Dialect triage needs LLM |
| Feasibility | 15% | 5 | Very buildable |
| Impact & scalability | 15% | 4 | Every flood RHU |
| Weighted total | 4.10 |