Swohz: AI TV Show Discoveries

Inspiration

Have you ever spent 40 minutes scrolling through streaming platforms, only to end up rewatching a show you’ve already seen five times?

Most modern recommendation engines push the same mainstream titles over and over again, keeping users trapped in filter bubbles. We built Swohz to break this endless loop by combining pure, randomized exploration with smart AI recommendations—helping people discover hidden gems, underrated masterpieces, and forgotten classics tailored to their taste.

What it does

Swohz is an AI-powered TV show discovery app designed to defeat choice paralysis:

Randomized Discovery Engine: Applies deep filters (genre, release year, minimum rating thresholds) to draw completely unbiased, random pages from the TV database so you always see fresh options.
AI Recommendation Feed: Analyzes saved favorites to identify thematic patterns, tones, and sub-genres, generating hyper-personalized suggestions.
Favorite Tracker & Watchlist: Lets users save shows with a single tap to continuously refine their AI recommendation engine.

How we built it

Mobile App: Built as an Android application with a focus on fast navigation and a streamlined, high-contrast UI.
Data Layer: Integrated with comprehensive TV show metadata APIs to query broad catalog indices and handle dynamic filter combinations.
Randomization Pipeline: Developed custom pagination logic to fetch randomized result pages within user-defined quality boundaries.
AI Engine: Designed a recommendation mechanism that maps user preferences and saved show vectors to suggest semantically similar series.

Challenges we ran into

Handling API Constraints: Standard media APIs aren't designed for true random sampling. We had to implement smart paging offsets and result balancing to avoid rate limits while ensuring genuine randomness.
Solving the Cold Start Problem: Making the AI provide accurate recommendations even when a user has only added 1–2 favorites required fine-tuning feature-matching heuristics.
Filtering Noise: Ensuring that randomized results surfaced high-quality, watchable content rather than low-rated noise required strict rating and metadata validation layer integrations.

Accomplishments that we're proud of

Built a working Android app published on the Google Play Store.
Created a functional discovery loop that successfully combines stochastic (random) search with deterministic (AI/Filter) preference matching.
Received positive early user feedback on breaking the "infinite scroll fatigue" on major streaming platforms.

What we learned

How to optimize user discovery flows to minimize decision fatigue in media consumption.
The balance between deterministic filtering (ratings, genres) and stochastic discovery mechanisms (pure randomness).
Best practices for designing lightweight, prompt/vector-driven recommendation layers for mobile clients.

What's next for Swohz: AI TV Show Discoveries

Streaming Provider Sync: Direct integration to show users exactly which platform (Netflix, HBO/Max, Prime, etc.) streams the show in their region.
Natural Language AI Prompts: Allowing users to search using complex vibes, such as "Show me a dark sci-fi thriller set in the 90s."
Group Discovery: A shared mode that merges two friends' favorite lists to find the perfect show to watch together.