Hire Me
All selected work

PER-001-CS-003 · Executive summary

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.

Context

Context

Patients needed clear triage guidance while clinicians needed concise, structured context instead of raw chat history.

Constraints

Constraints

Clinical safety boundaries, bilingual input, authenticated roles, incomplete patient context, and a fixed demonstration timeline all had to remain explicit.

Combine multilingual conversation, RAG, clinical safety, vision, authentication, and handoff workflows without presenting the assistant as a diagnostic tool.

Decision

Decision

  • 01Keep safety and red-flag escalation outside free-form generation.
  • 02Ground answers with retrieval and verify high-risk output after generation.
System

System

Implementation boundary

  • FastAPI
  • Next.js
  • PostgreSQL
  • pgvector
  • Redis
  • Docker
  • RAG
  • LLM providers
Reliability & contribution

Reliability & contribution

Specific ownership and controls—not a claim over the entire system.

  • 01Contributed to a bilingual clinical assistant in a team setting with safety-oriented workflows.
  • 02Repository artifacts document backend, retrieval, and interface components; individual responsibilities remain under review.
  • 03Team-project materials include architecture, demonstration, deployment, and fallback documentation.
Verification

Verification

Verified by approved artifacts

Repository artifacts and acceptance checks cover bilingual routing, safety escalation, authenticated handoff, and documented deployment paths; no clinical-efficacy claim is made.

Outcome

Outcome

  • Bilingual patient, doctor, and admin flows prepared for a live defense.
  • Production-style Docker, observability, security, and test coverage documented in the project.

08

Privacy

No patient records, private credentials, internal endpoints, or identifying clinical screenshots are published; the case shows only the engineering boundary.

Related project

Have a similar operational boundary?

Start with the problem, existing process, risk, and the result that must be verified.

Discuss a related system