Projects
AI Fitness Assistant
AI Fitness Assistant generates tailored workout plans and provides guidance based on goals and feedback. The core problem was building a flexible prompt system that adapts routines while maintaining user safety. The current focus is stabilizing the planner and preparing a production-ready release.
The problem this solves
Generic workout plans don't account for the user's current fitness level, available equipment, time constraints, or recovery needs. Personal trainers solve this but aren't accessible to most people. This project uses AI to generate plans that adapt to the user's actual situation — not a template designed for everyone.
What it does
- Onboarding flow to capture fitness level, goals, and equipment
- AI-generated personalized weekly workout plans
- Exercise swap suggestions when equipment isn't available
- Progress logging with weekly check-ins
- Adaptive plan updates based on logged feedback
- Rest day and recovery recommendations built into the plan
Hard parts and how I solved them
Fitness advice done wrong can cause injury. The prompt engineering had to be careful about what the model was allowed to recommend — especially for users with injury history or health constraints. I spent time building safety rails into the generation prompts so the output stayed responsible. The second challenge was making adaptation feel natural: when a user logs that a workout was too hard, the next plan should adjust meaningfully without completely discarding the structure.
Tech stack
Outcome
A working fitness assistant that generates realistic, personalized plans and adapts them based on user feedback. The onboarding-to-first-plan flow is under two minutes, which keeps drop-off low before users see the value.
My role
Solo — product design, AI prompt engineering, safety guardrails, plan adaptation logic, and the Next.js frontend.