RAG system for biomedical literature search. Combines dense + sparse retrieval, cross-encoder reranking, query expansion, and conversational memory.

Complete end-to-end MLOps pipeline with Apache Airflow orchestration, MariaDB ColumnStore OLAP, MLflow experiment tracking, Redis inter-task storage, Great Expectations data validation, and FastAPI model serving with drift detection.

Distributed system for managing ML training pipelines across GPU clusters with automatic checkpointing and fault tolerance.

Event processing engine handling 100K+ events/sec with exactly-once semantics for production data pipelines.

Full-stack dashboard for tracking ML experiments, comparing runs, and visualizing metrics with team collaboration.