Projects
A showcase of my technical projects, demonstrating my ability to build innovative solutions across web, AI/ML, and DevOps.
Published RAID (Rapid Automated Interpretability Datasets), a Python framework for interpreting how LLMs process and understand source code. Built neural activation analysis, AST-based code analysis with tree-sitter, and multi-granular token labeling to map transformer behavior to code structures.
Technologies Used:
Created a 2-player Gomoku game featuring an AI opponent powered by alpha-beta pruning and handcrafted evaluation functions, integrated within a modular game engine that optimizes search depth and pruning to enable fast, near real-time decision-making and strong competitive play against human opponents.
Technologies Used:
Developed an image forum web app that allowed users to upload, download, search, and categorize images, deployed on AWS (EC2, RDS, S3) and later migrated to Google Cloud Platform, gaining hands-on experience across multiple cloud environments.
Technologies Used:
Built a containerized MERN e-commerce application featuring a unique avatar-based navigation UI. Implemented payment, delivery, and cart features along with a REST API for admin product management. Deployed database on MongoDB Atlas Cloud, backend on Render, and frontend on Vercel.
Technologies Used:
Led front-end development of SwipeHire, a Tinder-style Android job portal app that makes the application and recruiting process more efficient. Implemented swipe-based profile browsing, real-time chat via WebSockets, and a GitLab CI/CD pipeline to streamline deployment.
Technologies Used:
Implemented A*, BFS, DFS and IDS graph algorithms with heuristics to solve various levels of the 8 puzzle game. Utilized three heuristic functions for A* to explore their impact on search performance. Conducted tests across various problem depths to analyze average run time and nodes explored.
Technologies Used:
Developed a real-time trading alert system that automatically computes classic pivot support/resistance levels from prior session data, monitors intraday price action and VWAP, and sends actionable notifications to Discord when price approaches key decision zones. Utilized pandas and NumPy for data processing.