Developed on a Research Project, spliDT ML-Framework, a partition-based approach for implementing decision tree inference directly in real-time networks at the data plane for better security and performance using P4. (Codebase could be private due to its research nature)

Extended a basic Layer 3 forwarding program with a simplified version of In-Band Network Telemetry (INT), which tracks the path of each packet and records queue lengths at every switch by attaching switch IDs and queue depths, which can be read at the destination.

Improved API Fetching Mechanism & Removed Hardcoded Tokens to enhance reliability and security of the application.

Developed a content-based Movie Recommendation System using Streamlit for deployment and Kaggles TMDB dataset for training, leveraging libraries like Pandas, and Scikit-learn to achieve ~85% relevancy accuracy in recommendations.