Trainr
AI-Powered Fitness Challenge Platform
A SaaS platform that lets fitness coaches design and sell personalised challenges to their clients, with AI generating tailored plans in seconds.
- Stack
- LaravelPHPVue.jsMySQLRedisAI/LLM APIsAWS
Hours → seconds
Plan setup time
Multi-tenant SaaS
Architecture
Queue-backed
Processing
The problem
Coaches were spending hours hand-building each client's challenge plan in spreadsheets, which capped how many clients they could take on and made progress tracking manual and error-prone.
What I built
I built a multi-tenant SaaS platform where coaches define challenge templates and AI expands them into personalised, day-by-day plans. Plan generation runs on a queued worker so the UI stays responsive, and a real-time progress pipeline records client activity as it happens.
The outcome
Setup time per client dropped from hours to seconds, letting coaches scale their roster without adding admin overhead. Coaches get a live view of every client's activity and performance.
Overview
Trainr is a SaaS platform enabling fitness coaches to design, personalise and sell challenges to their clients — with AI-driven automation doing the heavy lifting on plan generation.
What I built
- AI plan generation — coaches describe a challenge; the system generates a tailored, structured plan in seconds instead of hours of manual setup.
- Queued generation pipeline — AI calls run as background jobs so slow upstream responses never block the request cycle.
- Real-time progress tracking — client activity and performance stream into a coach-facing dashboard.
- Multi-tenant billing — each coach runs their own storefront of challenges under one platform.
Engineering notes
The interesting constraint was latency variance in AI responses. Generation was moved entirely behind a Redis queue with idempotent job handling, so a retry never produces a duplicate plan. Progress events are written on a separate write path and aggregated for the dashboard, keeping the hot read path cheap.