Medical imaging platform for AI-based virtual staining of histopathology slides. Owned the architecture, pipeline design and module delivery.
- Designed a large-file ingestion pipeline for 20–40 GB uploads using AWS S3 multipart upload with pre-signed URLs, removing client-side timeouts.
- A DAG-based validation job runs after upload; results stream through Redis Streams to Celery workers that parse metadata such as file type, species, organ and diagnosis.
- Built a configurable rule engine that reads the parsed metadata, picks the right stainer and routes jobs to AI inference.
- Also built the analytics dashboard, rule configuration UI, settings panel and a central upload manager.