HeeyCoach | Building an AI-Powered Coaching Platform

HeeyCoach is a platform that helps football coaches plan, generate, and manage training sessions using structured data and AI.

The goal was to move beyond generic AI outputs and build a system that produces context-aware, realistic training sessions based on real coaching constraints (age group, session phase, objectives, etc.).

AI generation pipeline: structured session output using schema-based responses (not free text), ensuring consistency and direct usability in the UI

Domain model: TrainingTask → TrainingSession → SessionTask, with rich metadata (technical, tactical, physical, cognitive) to support filtering and extensibility

Context filtering: dynamic selection of relevant exercises based on constraints (age group, phase, objectives), avoiding prompt overload and improving output quality

Session builder: hybrid system combining AI-generated sessions with manual editing, preserving control while enabling automation

Application architecture: routing, data flow, and mutation boundaries using React Router data APIs, with clear separation of concerns

Multi-tenant system: club-based structure with teams, staff roles, and permissions, designed for scalability from early stages

Billing system: Stripe integration with club-level subscriptions, handling upgrades, downgrades, and lifecycle events

Infrastructure: React + TypeScript application, Prisma + SQLite with LiteFS replication, deployed on Fly.io with primary/replica awareness