Python Backend Developer | Building No-Bloat, High-Performance Systems

Saakaar Barthwal  • Peerlist

• A machine learning model that classifies yoga poses performed by the user in real-time from a live video feed 
based on a novel approach I came up with i.e. coordinate-driven classification.
• The model has an F1 score of 0.98 and a real-world accuracy of 98%.
• My paper comparing this model to other methods of pose classification has been accepted in the International 
Conference on Electrical Electronics and Computing Technologies (ICEECT) by IEEE.
• Built my own dataset containing tens of thousands of dataframes, for the training and validation of the 
model.

Realtime Yoga Pose Classification : A Novel Approach

• Designed, developed and deployed a high performance URL shortening service, having response time 
less than Bit.ly and TinyURL.
• Developed a custom algorithm for shortening URLs into a 5-character alpha-numeric phrase. The app also 
has support for user-defined custom URLs.
• The application is fully API-based, enabling easy integration with external systems and services.
• Implemented Google OAuth to get user data for easy sign-in.