
Sadeed
Led the "AI Innovators" team to build a production-grade, zero-hallucination AI co-pilot for Gastric Cancer clinical decision support. Designed to solve information overload for clinicians by providing source-cited, verifiable medical answers. Technical Highlights: • AI Architecture: Architected an 8-node state machine using LangGraph, incorporating query classification and confidence gating. • Pipeline Flow: Engineered a sequential AI pipeline leveraging hybrid retrieval (pgvector + BM25 with RRF) for precise context matching, Gemini for grounded generation, and an NLI faithfulness checker to strip contradicted claims. • Backend & Ops: Engineered a robust FastAPI backend with SSE streaming, utilized LlamaParse for complex medical document ingestion, and implemented LangSmith for full pipeline tracing. • Evaluation: Evaluated system performance and medical safety using RAGAS metrics.















