Trainerly
An AI personal-training application built around a person’s goals, equipment, training history and feedback. Conversational coaching and guided workouts bring the plan into day-to-day use.
We conceived the product and took it through requirements, user experience, architecture, implementation and deployment to private beta. The application uses Next.js and React on Vercel, Supabase for its database and authentication, Stripe for billing integration and Strava for activity data.
We designed the AI workflows and exercise-retrieval approach, and directed AI-assisted implementation using Claude Code against our own specifications. Product decisions, integration, review, validation and operation remain our responsibility.
- NEXT.JS
- REACT
- VERCEL
- SUPABASE
- POSTGRESQL
- STRIPE
- STRAVA
- N8N
AI that can be inspected
An administration console brings together workflow execution, versioned prompts and output evaluation. This supports investigation and comparison as the product evolves.
Visibility of usage and cost
Model calls are recorded with token usage and cost information. Rate limits, content controls and administrative review support the operation of the beta.
A controlled release process
GitHub pull requests create Vercel preview deployments and Supabase database branches, alongside automated code review. This provides a separate environment for validating changes before a production release.
Developing our own product means taking responsibility for the decisions around the code as well as the implementation: what to build, how to validate it, what it costs to operate and how to improve it. Trainerly brings that product-owner perspective into our client work, alongside direct experience of integrating and operating AI features.