Projects
Design Team
EcoCar EV Team at Virgina Tech (HEVT)
Leading the development of high-level autonomous capabilities for the 2023 Cadillac LYRIQ in the EcoCar EV Challenge. This includes features like auto lane change, eco-adaptive cruise control, and vehicle-to-vehicle communication. The project utilizes motion planning, perception, and a ROS2-based software stack to interface with the CAN bus for vehicle control, running on a custom compute platform with a React-based web interface for monitoring and control.
- PythonROS2CAN protocolDockerC++Reactsocket (TCP/IP) programmingVPN
Client Project
Agriculture Insurance LLM
Led the development of an interactive conversational assistant to simplify agricultural insurance decisions. Built using a Large Language Model (LLM) trained on Rainfall Index Insurance Standards and USDA data, the assistant offers clear, personalized guidance and visualizations. The project includes a user-friendly chat interface, a Flask API backend, and deployment on Virginia Tech's Endeavour cluster, showcasing the power of AI in demystifying complex insurance options.
- ReactNextJSPythonFlaskRedisChromaDB (Vector Database)DockerKubernetesLLMRetrieval Augmented Generationllama2Ollamagpt-4
Personal Projects
Youtube Q-A Chatbot Chrome Extension
Created a Chrome extension that lets users ask questions about YouTube videos in real-time. It uses a Retrieval-Augmented Generation (RAG) model to answer with evidence and timestamps from the video.
- ReactFlaskllama3.1RedisDockerPython
Hackathons
Powered Down
Developed a multi-user, third-party compatible smart home lighting system at a 36 hour hackathon. Using GPS to track user location, it automatically controls lights and integrates with smart accessories from different companies.
Prize Winner- ReactPythonFlaskNoSQL DBFirebaseGoogle Cloud PlatformTypescript
The Narrator
A 24-hour hackathon where my team and I built a website that transforms stories from narration or text into shareable virtual picture books with live narration in multiple languages.
- OpenAI Dalle ModelReactPythonFlaskMongoDBFirebaseGoogle Cloud PlatformTypescript
Food Snap
Developed an app, at a 24 hour hackathon, that uses a vision model and machine learning to generate recipes from user-taken food images. Users take a picture of their ingredients, and the app identifies the food, generates recipe options, and enables swipe-based selection. Liked recipes are saved to a dashboard for easy access.
Prize Winner- ReactNode.jsExpressMongoDBHeroku
Application Stack
A 24-hour hackathon where my team and I built a discord bot that keep track of job applications.
Prize Winner- ReactNode.jsExpressMongoDBHeroku