ML-powered restaurant recommendation engine — in development
- Sector
- Food Tech
- Location
- Remote
- Duration
- Ongoing
- Year
- 2024
- Services
- App developmentWeb developmentReactPythonFastAPIscikit-learn
- Recommendation response time
- < 200msRecommendation response timeFastAPI benchmarks
- Pilot cities in testing
- 2Pilot cities in testingCurrent rollout
- Collab + content-based model
- HybridCollab + content-based modelArchitecture decision
SmartBite is a product in active development that uses machine learning models to recommend restaurants to users based on cuisine preferences, dietary restrictions, location, budget, and past order behaviour.
Existing restaurant discovery apps rely on aggregate ratings that tell you a place is popular, not whether it matches what you actually want right now. A 4.8-star kebab house is useless if you are looking for a quiet brunch spot with vegetarian options.
Recommendation quality degrades at the edges — new users with no history, and niche preferences that don't fit broad categories — which is precisely where a genuinely useful recommendation engine needs to work well.
The recommendation layer combines collaborative filtering for users with sufficient history and a content-based fallback for cold-start users, using cuisine tags, price band, noise level, and dietary flags as feature inputs.
A FastAPI backend serves recommendations with sub-200ms response times. The frontend collects preference signals progressively — no forced onboarding survey — so the model improves as the user browses naturally.
The product is being built with real restaurant data from two pilot cities before any public launch, so the model has enough signal density to give meaningful results from day one.
SmartBite is currently in development. The recommendation model is in testing with a closed group of users across two cities. Public launch is planned once precision scores meet the internal threshold.
