◼ THE DEV AGENCY

mobile / 2025 / live

Bet.AI

Scan any bet slip. Get instant AI analysis.

ExpoReact NativeExpo RouterFirebaseRevenueCatOpenAIVertex AITypeScript

◼ Walkthroughs

◼ Live demo

Architecture

Expo SDK 52 frontend with a Firebase backend (Auth, Firestore, Storage, ~7K LOC of Cloud Functions). Six sports data providers feed parallel pipelines covering NBA, NCAA basketball, NFL, NCAA football, all major EU soccer leagues, MMA/UFC, and tennis Grand Slams. A custom CatBoost model deployed on Google Cloud Vertex AI computes 88 engineered features per prediction, hitting 70%+ accuracy on high-confidence interval player prop calls.

Features

  • Bet slip scan

    Image captured via camera or gallery, OCR'd by GPT-4 Vision, then parsed into typed bet slip data and rendered as a structured analysis.

  • Six-source data aggregation

    Parallel pipelines pull from The Odds API, SportsGameOdds (with key rotation), API-Sports, StatPal, API-Tennis, and WeatherAPI per request.

  • Custom ML props model

    CatBoost on Google Cloud Vertex AI. 88 engineered features per prediction from real-time player game logs. NBA and soccer props hit 70%+ accuracy on high-confidence interval calls.

  • Multi-sport coverage

    NBA + NCAA basketball, NFL + NCAA football, all major EU soccer leagues, MMA/UFC, and tennis Grand Slams + main tournaments.

  • Match intelligence layer

    Recent form, head-to-head, momentum, x-factors, and a written AI breakdown surfaced alongside odds and value ratings on every analysis.

  • Real-time orchestration

    ~7K LOC of Firebase Cloud Functions running parallel async calls per analysis. Firestore cache layer with TTL controls keeps response time low across six external APIs.