AI SOFTWARE ENGINEERING · NEUMARKER
Agentic medical assistant with verifiable retrieval
Built a Python-based assistant that searches patient records and public medical sources, then produces evidence-grounded outputs with human approval at high-impact steps.
- 01Combined MedCPT semantic search and BM25 keyword retrieval with Reciprocal Rank Fusion.
- 02Added human-in-the-loop approval before search, assessment, and treatment-related actions.
- 03Evaluated retrieval using patient identifiers, missing-record tests, and source-level citations.
semantic
keyword