100K+ users · +28% engagement
Fayvo
Social app with ML-powered recommendations
- React
- Redux
- TypeScript
- Django REST
- PostgreSQL
- AWS
Challenge
Grow engagement on a social network of 100K+ users where the default chronological feed was failing to surface relevant content.
What I built
- Led full-stack development across React/Redux/TS frontend and DRF backend
- Designed an ML recommendation system using neural networks over PostgreSQL features
- Optimized complex queries, cutting response times by 50%
Key decisions
- Chose neural collaborative filtering over simple heuristics for personalization quality
- Kept recommendations server-side so the mobile client stayed thin
Outcomes
- 100K+ users, 1K+ DAU
- +28% user engagement after recsys rollout
- 50% reduction in API response times