The client
My Workout AI is an AI-powered fitness and movement platform built on a philosophy of functional movement over traditional muscle isolation. Instead of generic workout plans, the platform builds fully personalised routines tailored to each user's body, goals, and movement capabilities — and adapts those routines in real time as the user progresses.
The challenge
The client had an extensive library of exercise content on their YouTube channel and a clear vision: an AI-driven fitness app that could intelligently recommend exercises from their own video library based on a user's individual profile. The core problem was connecting three complex layers together — user profiling (BMI, fitness goals, physical limitations), AI reasoning (interpreting that profile and recommending appropriate exercises), and content delivery (surfacing the right YouTube videos at the right moment in a user's plan).
- No AI integration — workout plans were static and generic
- YouTube content library was disconnected from the user experience
- No personalisation based on BMI, fitness level, or stated goals
- No mechanism to prevent plateaus — plans didn't adapt over time
- No iOS or Android apps — the experience existed only on web
Our solution
Custom GPT training: We fine-tuned and prompted a GPT model on the client's exercise library, movement philosophy, and fitness methodology. The AI understands the client's specific exercise taxonomy — movement patterns, difficulty levels, modalities — and recommends exercises in the client's own voice and framework, not generic fitness advice.
AI-powered onboarding and profiling: A conversational onboarding flow collects the user's BMI, current fitness level, physical limitations, available equipment, and fitness goals. This profile feeds directly into the AI recommendation engine, which generates a fully personalised workout plan referencing specific videos from the client's YouTube channel.
Dynamic plan adaptation: The platform tracks user progress — completed sessions, performance metrics, feedback — and the AI adjusts the plan in real time. If a user is progressing faster than expected, difficulty increases. If they're struggling, the plan adapts. Plateaus are algorithmically prevented by continuously varying stimulus.
Three movement modalities: The platform covers strength training, cardio, flexibility and mobility work, and mindfulness guidance — all delivered through the client's YouTube content, organised into structured progressive programmes by the AI.
iOS and Android apps: We built native-quality cross-platform iOS and Android apps using React Native, delivering the full AI-personalised experience on mobile. Users receive push notifications for scheduled workouts, track sessions in-app, and access their YouTube-linked exercise library without leaving the platform.
Results
- AI workout personalisation live on iOS and Android — App Store and Google Play
- GPT model trained on client's exercise library and movement philosophy
- YouTube content library fully integrated — exercises link directly to client's videos
- Dynamic plan adaptation prevents plateaus and maintains user engagement
- Three modalities (strength, cardio, flexibility, mindfulness) available at launch