Hybrid Recommender API on GCP (Chilldays)

The problem

ChillDays is a leisure app: it suggests restaurants to eat at, beach parties to join and night clubs for a night out. The client had a team of AI engineers but needed someone to lead them through building a hybrid recommender system and getting it into production, rather than another pair of hands writing code in isolation.

What I built

As consultant and technical lead, I guided the client's team through the design, review and deployment of the system:

• A hybrid recommender in Python that combines rule-based, collaborative and content-based filtering, so recommendations do not depend on a single signal.
• A REST API with several endpoints, each answering a different kind of request: where to eat, which beach party to go to, which club to pick.
• Data access against Firebase, where the app's data lives.
• Deployment of the API on Google Cloud Platform.
• Code review and technical mentoring for the engineers doing the implementation.

How it was delivered

My role was consulting and guidance rather than sole implementation. I set the architecture for the hybrid filtering, reviewed the team's code as it landed, and steered the GCP deployment so the recommender shipped as a service the app could call.

Outcome

A hybrid recommender deployed as a REST API on GCP, with endpoints serving restaurant, beach-party and night-club recommendations from Firebase data. The client's own engineers built it and own it; my contribution was the technical leadership that took it from design through review and into production.