Back to projects
Clinical RAG Evaluation Matrix
Interactive TypeScript evaluation framework and traceability matrix comparing local medical LLMs, vector retrieval and explainability overlays on chest X-ray reports.
- Project
- Open source
- Role
- Creator
- Period
- 2026
[ Overview ]
The problem
Comparing LLMs and retrieval setups for medical reporting is hard to do consistently without a shared view of results.
What I built
Built an interactive evaluation framework that links each configuration (model, retrieval and explainability overlay) to its results in one matrix.
[ Highlights ]
- Benchmarks local medical LLMs side by side
- Compares vector retrieval configurations backed by Qdrant
- Traceability matrix linking configurations to report outcomes and explainability overlays
[ Stack ]
- TypeScript
- Qdrant
- LLM evaluation