LumoroX
An AI-native movie discovery platform, evolved from the Movie Recommendation System.

Goals
- Give the SVD recommender a real product surface beyond an offline top-10 list.
- Support discovery by mood and natural language, not only by rating history.
- Explain why a title is recommended instead of returning an opaque ranking.
- Keep search and browsing instant on a large catalogue.
Approach & architecture
- 1
Recommender as a signal
The matrix-factorization model became one input among several rather than the whole system.
- 2
Catalogue search
Instant search over the full catalogue with trending rails for cold-start users who have no rating history.
- 3
LLM layer
Mood prompts and watch history feed an LLM layer that ranks candidates and writes a short reason for each suggestion.
- 4
Watchlists
Per-user watchlists on Supabase feed taste signals straight back into the recommendation inputs.
- 5
Cinematic UI
A glassmorphic, motion-rich interface keeps browsing closer to a streaming app than a data tool.

Tech stack

Results
RMSE improvement inherited from the SVD foundation
shipped at lumorox.lovable.app
search, trending, watchlist and AI chat in one surface
Frequently asked questions
More case studies
J.A.R.V.I.S — NIX
A local-first real-time voice AI that can hear, see and control your machine.
LumoroX Park
A live driveway-parking marketplace, evolved from the Parking Space Management System.
Telecom Customer Churn Prediction
An end-to-end ML pipeline that flags at-risk telecom subscribers before they leave.