Projects

Built an empathetic AI for self-reflection using LangChain and HeyGen for natural conversations. Enabled real-time memory with ChromaDB for personalized insights and pattern recognition. Developed with Python, Flask, React, and AMD Minisforum AI 370 for NPU+GPU compute.

Built a 1/18-scale autonomous vehicle using ROS 2 Humble on Raspberry Pi 4, navigating via RGB camera-based blue line following and LiDAR odometry. Applied HSV thresholding and contour detection with OpenCV for traffic sign detection, achieving ~80% stop sign accuracy, and implemented EKF SLAM for robust localization. Designed a modular ROS 2 system with nodes for camera processing, line detection, PID control, and slam_toolbox for real-time mapping and Tkinter-based telemetry visualization.

Successfully reproduced the results from the "Look Listen & Learn" paper, implementing the L3Net model for audio-visual correspondence tasks using few-shot learning techniques.

Built an end-to-end interpreter that takes Python-like syntax, implementing features such as parsing, control flow, hashes, garbage collection, and superclasses.

Developed an array of non-evasive sensors for collision detection in a helmet with end-to-end support for immediate ambulance alert and injury profile using ESP32 Kit.