I created medeval-framework, an open-source suite for evaluating clinical and high-stakes LLMs. It measures calibration, semantic drift, and hallucination penalties, with 170+ unit tests and 100% test coverage.
For my SickleCellDomain-SpecificAI&RAGPipeline project, I fine-tuned an open-weight model on curated medical corpora and integrated a ChromaDB-backed RAG pipeline to ground responses in clinical contexts. I also set up experiment tracking with MLflow via DagsHub and live inference hosting on Hugging Face.
At GracetechGroupofCompanies, I engineer production web applications and backend architectures using Django and FastAPI, integrating React and TypeScript interfaces with backend data services and ML workflows. Previously, I worked on AI agent evaluation, backend and data engineering, and machine learning projects.
