The research paper describes the project on developing LLaMA models to enhance language model efficiency. I also implemented the transformer-based architecture with RMSNorm and rotary embeddings for consistent training. The LLaMA-2 model uses grouped-query attention, improving scalability and performance. Despite challenges with Apple Silicon, I fine-tuned the model using LoRA for efficient training and better NLP task performance.

I developed a CLI tool designed to analyze YouTube videos by fetching their total duration and identifying the minutes dedicated to sponsorships. This tool was created to address the frustration of encountering lengthy sponsored content.

I developed an RFID Scanner using FRDM K66, ESP32, and Flask (Python). The system includes an MFRC522 RFID scanner and a DS1302 Real-Time Clock for precise timekeeping. A server at the backend handles operations with data stored in SQLite, and an auditory buzzer confirms successful check-ins. This project streamlines attendance tracking and provides event organizers with actionable insights.

I implemented an Embedded Debug Logging System to enhance debugging capabilities. The system sends logs back to a central server from the app where the debugger is deployed. Log messages are descriptive and concise, including the event details, timestamp, log severity, function name, and line number. This improves the efficiency and accuracy of debugging processes.

This is not entirely a project but rather an idea that got me and my team to win the 4th place at Seneca Hackathon, The project was Denvide and was aimed to create a platform that would suffice both the requirements for the Students and the Landlords, This was an impact driven idea that me and my team came up with in 3 days