Book Recommendation System
2025-08 Production-grade hybrid recommender with warm/cold support, real-time similarity search, and daily retraining with hot reloads.
I build production-ready backends and agentic AI systems that combine retrieval, tool use, and orchestration with solid data engineering. My business + IT background (HEC Montréal, bilingual program) helps connect technical delivery to product outcomes.
Recently, I shipped an agentic workflow on top of a FastAPI stack: retrieval-augmented generation, vector search, and tool-calling (search, internal ML services, data lookups), all behind authenticated endpoints with observability and safe fallbacks. Earlier work includes an end-to-end book platform (normalized SQL, embedding pipelines, and automated training/rollouts), which I now leverage as one of several tools inside the agent system.
Currently exploring: robust agent design (planning, tool routing, memory) and O’Reilly’s Data Engineering Design Patterns to deepen pipeline reliability, lineage, and deployment practices.
RAG pipelines with tool-calling: web/data search, internal ML tools, and guarded execution with structured prompts, schemas, and evaluation.
FastAPI services with caching, pagination, auth, and observability; deployed behind Nginx with safe hot reloads.
SQL modeling, feature prep, scheduled jobs, artifact promotion, and environment-aware rollouts.