Cinekholm
AI-powered movie tracking platform with iOS & Android apps.
Cinekholm is an AI-powered movie tracking and discovery platform available on web, iOS, and Android. It combines Angular (TypeScript) with SSR on the frontend and a Node.js/Express backend using MongoDB, Redis, and BullMQ for scalable processing. The product includes multidimensional taste profiling, personalized recommendation workflows, Movie Night group picks, and real-time social interactions supported by Socket.IO.
Context
Picking a film with friends kept going the same way: the bigger the group, the longer it took. In my own friend groups I watched it stretch to 40–50 minutes, and the night often ended with a film nobody really wanted, or with no film at all. Before building anything I went through the research on it. Nielsen's 2025 State of Play puts the average at 14 minutes of searching per session for one person, and the choice-overload literature, from Iyengar & Lepper's jam study to Schwartz's The Paradox of Choice, shows that more options make people less likely to choose at all. Streaming catalogs only keep growing.
Problem
A longer list doesn't help someone who is already lost in one. The app had to learn a person's taste from a handful of quick ratings, turn it into recommendations they would actually like, and do the harder version of the same thing for a group: find one film that works for everyone in the room, fast, without anyone feeling outvoted.
Approach
I built the whole product on my own: the Angular SSR web app, the iOS and Android apps and the Node.js backend. Every title is scored by GPT-6 Luna on ten mood dimensions such as suspense, intimacy and gravitas. A user's ratings, weighted toward recent ones, add up to their cinematic profile, and an in-app assistant running on GPT-6 Luna answers in conversation when someone would rather ask than browse. Movie Night combines the profiles of everyone in the group into a single short deck; everyone swipes on the same cards, and a film counts as a match once at least 70% like it (everyone, in groups of two or three). The heavy work runs in BullMQ queues with Redis caching, so building a group's deck happens in the background instead of keeping the phone waiting.
Outcome
Close to 400 users so far, with a full five-star rating on both the App Store and Google Play. The apps are live and available to download on both stores.