AI Recommendation Engine Integration

I’m kicking off a project that needs a compact, production-ready recommendation engine powered by modern LLM services. The stack is firmly in Python land—pandas/numpy for data wrangling and either FastAPI or Flask for the lightweight REST layer. You’ll be orchestrating calls to OpenAI, Claude and Gemini, blending their strengths to return smart, ranked suggestions in real time.

Here’s the immediate scope: turn our raw interaction logs into features, design a prompt or chain that yields high-quality recommendations, wrap the logic behind clean endpoints, and ship a small Dockerised service I can drop into our infra. Time is critical; I want the first usable version live as soon as you can manage.

Deliverables
• A Python package that ingests a CSV/Parquet dataset, produces feature vectors, and queries the LLM APIs to generate recommendations
• REST endpoints (FastAPI/Flask) for /recommend and /health, covered by basic tests
• Dockerfile and minimal deployment guide

Acceptance criteria
– Given a user ID the API returns the top-N items in <500 ms on modest hardware
– Unit tests pass in CI and endpoints are documented via OpenAPI spec
– All external calls work with OpenAI, Claude and Gemini keys provided at runtime

There’s plenty more we could build—including user-level personalisation and A/B experimentation—so a smooth first milestone can grow into a longer engagement. If this sounds like your wheelhouse, let’s move fast.