FormCoach

Form Coach gives feedback on video upload of Jacob doing a push up
Meal suggestions from Nutrition Agent
Coach Chat Bot with TTS
Nutrition Dashboard
Dashboard

Inspiration

We noticed a pattern at HackUSF: people want to get healthier, but they get stuck on the basics “Am I doing this exercise right?” and “What should I do next?” especially when they don’t have a coach or gym buddy. We wanted to build something that makes good form and consistent progress feel more accessible, like having a calm, supportive trainer on demand.

What it does

FormCoach helps users plan and improve workouts by giving coaching-style guidance focused on proper technique, safer training, and sustainable routines. It turns fitness goals into actionable steps, helps users stay consistent, and provides feedback that feels like a real coach: clear, encouraging, and practical.

How we built it

We built FormCoach as an AI-powered coaching experience with an “agent” approach: different parts of the system handle planning, guidance, and adapting suggestions to the user’s goals and constraints. We connected the orchestration layer to prompt-driven coaching logic, organized the project for quick iteration, and tested flows end-to-end to make sure the experience stayed simple and helpful.

Challenges we ran into

The hardest part was balancing accuracy, safety, and clarity. Fitness advice can’t be vague, but it also can’t be overly confident or risky. We also had to keep responses consistent across different user goals, avoid information overload, and make sure the “coach” voice stayed supportive instead of robotic.

Accomplishments that we're proud of

A cohesive coaching narrative that feels like guidance, not just answers
A structured, agent-based design that’s easy to expand
Clear user-facing flows that make the project feel like a real product, not a demo
Thoughtful constraints around safer recommendations and realistic progress

What we learned

We learned that building a “coach” is as much UX as it is AI: tone, structure, and the order of information matter. We also learned the value of modular agent orchestration—once the pieces are separated cleanly, improvements become faster and more reliable.

What's next for FormCoach

Next, we want to make FormCoach more personalized and measurable: user profiles, progress tracking, routine periodization, and deeper feedback loops. We also want to expand into richer inputs (like exercise clips or sensor data), add safer guardrails for edge cases (injuries, limitations), and turn the prototype into a polished app experience with onboarding, saved plans, and shareable progress.