← Back to Portfolio

SportsCove

Live coaching marketplace on iOS, Android, and web — verified coaches, live video, and AI practice between sessions.

Tech Stack

FlutterRiverpodSupabase (PostgreSQL)LiveKit EdgeStripe ConnectRazorpay RouteDeno Edge FunctionsAWS S3Redis

The Problem Space

The USD $40B+ amateur sports coaching industry suffers from extreme fragmentation, archaic discovery, and zero cross-border payment infrastructure. Elite coaches face a ceiling on their local reach and lack tools to monetize globally, while athletes face prohibitively high costs ($1,000+/year) for structured guidance. The mandate was to build a zero-latency, high-concurrency platform capable of routing synchronous video streams and asynchronous escrow payments across 150+ countries without breaking compliance or user experience.

The Solution

SportsCove was architected from the ground up as an AI-powered aggregator. For the demand side (athletes), it functions as an intelligent discovery engine with automated matchmaking and structured progress loops. For the supply side (coaches), it is a full-stack SaaS operating system—handling dynamic scheduling, multi-currency escrow payouts, automated tax compliance (including TDS for India via Razorpay Route), and CRM. Phase 2 integrates an AI Virtual Trainer utilizing computer vision for real-time biomechanical analysis.

Technical Architecture & Components

The platform employs a robust microservices architecture. The frontend is a highly modular Flutter monorepo (iOS, Android, Web) utilizing Riverpod for reactive state management. The backend relies on Supabase (PostgreSQL) for transactional integrity, augmented by Deno Edge Functions deployed globally to minimize cold starts. Real-time synchronous video sessions are managed via LiveKit, dynamically scaling nodes based on geographic density. A custom UnifiedPaymentOrchestrator securely routes transactions between Stripe Connect (global) and Razorpay (India), while asynchronous background jobs (ledger reconciliation, notifications) are queued via Redis and executed by serverless workers.