All selected work
Healthcare AI · Team project
Bilingual Clinical AI Triage Platform
A safety-gated Arabic and English patient assistant that grounds responses in a clinical knowledge base and produces structured doctor handoffs.
system pathRAG + red flags
Patient
Safety-gated AI
Doctor handoff
Context
Patients needed clear triage guidance while clinicians needed concise, structured context instead of raw chat history.
Engineering challenge
Combine multilingual conversation, RAG, clinical safety, vision, authentication, and handoff workflows without presenting the assistant as a diagnostic tool.
01
My contribution
Specific ownership and delivery work—not a claim over the whole system.
- 01Hardened runtime, provider routing, authentication persistence, and the patient-to-doctor handoff chain.
- 02Aligned backend schemas, API payloads, frontend types, and tests for readable clinical handoffs.
- 03Prepared bilingual architecture, demo, deployment, and fallback documentation.
02
Key decisions
- 01Keep safety and red-flag escalation outside free-form generation.
- 02Ground answers with retrieval and verify high-risk output after generation.
03
Verified outcomes
- Bilingual patient, doctor, and admin flows prepared for a live defense.
- Production-style Docker, observability, security, and test coverage documented in the project.
System stack
- FastAPI
- Next.js
- PostgreSQL
- pgvector
- Redis
- Docker
- RAG
- LLM providers