AI Health Chat
Multilingual health-information assistant on web and WhatsApp
- Status
- Public beta
- Engagement
- Independent product
- Role
- Founder · full-stack & AI engineer
The problem
Multilingual health information is hard to access reliably, especially on low-bandwidth connections and messaging platforms.
What was built
A live multilingual AI health-information assistant on the web and Meta WhatsApp Cloud API, offering educational triage support, multimodal image understanding, vital-sign tracking and physician-ready consultation summaries.
Architecture and engineering decisions
A React and TypeScript client backed by Supabase for data and access control, with server-side LLM provider integration and a WhatsApp Cloud API delivery path. Privacy controls and a published medical disclaimer sit around the AI workflow.
Audience
People seeking educational health information in English, Kiswahili, French, Spanish or Dutch via web or WhatsApp.
Technical highlights
- Multilingual LLM pipeline across 5 languages
- Web and WhatsApp (Meta Cloud API) delivery
- Multimodal image understanding
- Longitudinal vital-sign tracking
- Structured physician-ready consultation summaries
- README-reported 144 passing tests
My responsibilities
- Product design
- Full-stack implementation
- LLM integration
- Testing and safety copy
Why it matters for a client
Shows how a sensitive AI workflow can ship with explicit safety boundaries, multilingual UX, messaging-channel delivery and human-readable outputs.
Constraints and trade-offs
Educational triage support only — not a medical diagnosis tool; no clinical validation is claimed; safety boundaries are stated on the platform.
Privacy and accessibility
Privacy controls and a published medical disclaimer; not a substitute for professional medical diagnosis or emergency treatment.
Verified evidence
Live at aihealthchat.qzz.io. The project README reports 144 passing tests (README-reported, not re-run for this site). 1st place at the Power Learn Project Class of 2024 graduate hackathon (December 2024).
Limitations
Not clinically validated; does not diagnose patients; hackathon-originated independent work, not commissioned or institutionally approved. Test counts are README-reported.