Ai-Fitness
Exercise recommendation
Tell it which muscles you want to work, your level and what equipment you have: the web app finds the right exercises in a dataset of thousands of entries.
- Type
- Tools and automation
- Year
- 2024
- What I did
- Whole project and deployment
- Links
- GitHub
The problem is concrete: there is an ocean of documented exercises and no simple way to pull out the ones that suit you today, with the equipment you actually have.
The app asks three things — beginner or advanced, target muscles, available equipment — and returns exercises with instructions and images. The match is not a filtered query: exercise descriptions are vectorised with TF-IDF and compared against preferences using cosine similarity, so near matches a strict filter would drop still come through.
Flask backend with six routes, data on MongoDB, front-end in HTML, CSS and JavaScript. The part that taught me most was putting it online: an EC2 instance, an IAM role to read credentials without writing them into the code, nginx in front of the app, a Let’s Encrypt certificate and a domain pointed with DNS. The instance is no longer running; the code is still on GitHub.