The goal - what I'm exploring
Net-Trailer started as a full-stack streaming clone and became my testbed for grafting AI onto an existing product. The thing I'm exploring: can "what should I watch?" feel like asking a person who actually knows your taste? So it pairs natural-language search with a recommendation engine that learns from your behavior - and explains why it picked each title.
How it uses AI
Net-Trailer's AI is a Gemini-powered recommendation layer over a TMDB catalog. Natural-language search sends your phrasing to Gemini, which extracts intent - media type, genres, mood - and turns it into a tailored row, behind a multi-model router that falls back automatically when a model is busy. A separate personalized engine builds a profile from your full watch and rating history, with a real-time feedback loop that reshapes future picks and a "why this was recommended" explanation attached to each title.
How it works
- Tell it what you’re in the mood for in plain English - a Gemini-powered analyzer reads your intent (media type, genres, vibe) and builds a matching suggestion row on the fly, with a multi-model router that falls back automatically.
- A personalized “Recommended For You” engine learns from your full watch and rating history with a real-time feedback loop - dislikes and skips reshape future picks, and each suggestion can explain why it surfaced.
- Browses live TMDB data - trending titles, genre rows, and trailer playback - with a cinematic, Netflix-style UI built in React and Tailwind CSS.
- Secure sign-in with NextAuth; sessions gate the watchlist and checkout flows.
- Full subscription checkout implemented twice over: both Stripe and PayPal payment flows, end to end.
- Your watchlist and profile persist in Firebase, syncing in real time across sessions.
- Server-rendered with Next.js and deployed on Vercel, with Cloudinary handling optimized image delivery.