2024-2025
CheckFit
I translated 14+ years of coaching judgment into an AI movement coach that adapts training and nutrition to real-life constraints.
Turned 14+ years of coaching judgment into product logic and a beta app.
Translated recovery, schedule, nutrition, and pain signals into decisions.
Generated and adjusted coaching plans from real user context.
What Was Built
CheckFit was an AI-powered movement coach built around a simple problem: most people do not need more fitness information. They need a plan that changes when their body, schedule, recovery, food, pain, and constraints change.
The beta translated messy user context into adaptive fitness, nutrition, and biomechanics coaching decisions. It included intake flows for goals, training history, constraints, and preferences; workout and nutrition plan generation; daily adjustment logic based on sleep, food, supplements, soreness, and schedule changes; and onboarding flows for early users.
My Role
I co-founded the product and built the first web application using Laravel, Filament, PHP, Tailwind CSS, and Anthropic's Claude API.
I translated 14+ years of fitness and coaching experience into product logic: intake structure, decision rules, prompt architecture, plan-generation flows, daily adjustment logic, and the product boundary between AI guidance and human coaching judgment.
Result
CheckFit produced a working beta that could generate and adjust training and nutrition guidance around real user inputs.
The product proved the core wedge: domain judgment that normally lives inside a coach's head could be converted into software that reduces planning burden, explains the next action, and adjusts when real life changes.
Why It Mattered
Most fitness plans assume the user is static. Real users are not. Sleep, soreness, stress, nutrition, medication, pain, motivation, equipment, and schedule constraints change constantly.
The opportunity was to use AI to make the plan adaptive without making the user become their own full-time analyst. The product had to turn changing inputs into fewer, clearer decisions.
What This Proves
I can take hard-earned domain expertise, turn it into product logic, and build a working AI health product from zero. The pattern is the same as the rest of my work: messy inputs in, fewer better decisions out.