Streaming ingestion, hybrid retrieval (dense + BM25), learned ranking, a RAG shopping assistant with citations, and Spark analytics over Amazon ESCI product data. Built on Qdrant, Redpanda, Postgres, S3, Redis, FastAPI, gRPC, LightGBM and PySpark. Deployed on AWS (EKS) with Terraform, not running in production.
Status: design phase; no service code yet.
uv sync # Python 3.12 + dependencies
uv run pytest # fast tests, no Docker
make up # Redpanda + Qdrant + Redis (needs Docker and GNU make)