Gymni - AI Workout Companion

Home Page
Exercise Page
Feedback Page

Inspiration

Many people struggle to maintain proper form during workouts, leading to injuries or poor results. Existing fitness apps offer fixed routines with little adaptability. We set out to build a smarter solution — an interactive, AI-powered workout companion that adapts in real time to any exercise and provides personalized feedback and guidance.

To support collaboration, we integrated a secure messaging system where users can share progress and feedback with friends or trainers. In response to a recent credential leak affecting over a billion users, our service uses a post-quantum cryptography (PQC) algorithm and includes a secure handshake mechanism that lets users change their private keys, ensuring encrypted, future-proof communication.

What it does

Gymni lets users choose any exercise — there's no fixed list. It uses:

MediaPipe to track full-body posture and count reps in real-time.
Gemini to configure pose parameters and provide feedback.
Voice Guidance to count reps and guide the user mid-session.
MongoDB Atlas to manage authentication, chat, and history.
PQC-encrypted chat to allow secure interaction between users.

How we built it

Frontend: HTML, CSS, and JS
Backend: Flask (Python)
Pose Detection: MediaPipe
Exercise Config & Feedback: Gemini API
Voice Interaction: Python TTS API
Database: MongoDB Atlas
Chat Security: Post-Quantum Cryptography (PQC) implementation

Challenges we ran into

Generating dynamic exercise configurations from natural language prompts.
Mapping Gemini's output into usable logic for real-time tracking.
Managing consistent rep counting across varied postures and users.
Implementing secure yet lightweight PQC-based chat.

Accomplishments that we're proud of

Created a fully dynamic, AI-driven fitness assistant from scratch.
Enabled users to do any exercise — no limits! We mean it. Any exercise!
Implemented real-time posture tracking, rep counting, voice interaction, and exercise history.
Built a secure chat system using PQC encryption.

What we learned

How to integrate LLMs for backend logic generation.
Advanced pose tracking and interpretation using MediaPipe.
Structuring full-stack apps with security and scalability in mind.
Basics of post-quantum cryptography and its application in secure communication.