Alaska_BIO-BLOCK_GSoC'26 (Suryansh Maurya) Project no.6

My proposal tackles the critical challenge of making raw medical data safe to share on Bio-Block without destroying its usefulness for machine learning. To achieve this, I'm building a "zero touch" FastAPI ingestion pipeline that acts as an automated watchdog, seamlessly using Microsoft Presidio to scrub sensitive personal health information from text, DICOM files, and medical images before they ever reach storage. Because we still need this data to be valuable for AI research, I'm implementing flexible "Privacy Profiles" that allow the system to intelligently preserve general demographics rather than just blindly deleting every identifier. Finally, since highly anonymized data is very difficult to search through, I'm building a hybrid search engine that combines exact keyword matching (BM25) with context aware semantic search (ChromaDB), ensuring researchers can easily discover the exact datasets they need without running into redundant search results.