Recommender From Scratch: Building, Monitoring, and Learning in Public

In this series, I’m building a real-time recommender system from scratch, starting locally with Docker Compose.

The goal? Build a full-stack pipeline: FastAPI + Redis or Faiss, with Spark for candidate generation and Prometheus + Grafana for observability.

Right now, I don’t have my homelab available — so the first step is to simulate everything locally while documenting every piece of learning along the way.

Expect posts about tradeoffs, debugging, architecture decisions, and real-time challenges.