Projects

Parkinson's Care VR System
github slides post
C# Unity TensorFlow React Flask SVM CNN
Georgetown University H2AI Hackathon: 1st Place Grand Prize & Patient Safety Award
  • Won the H2AI Grand Prize ($6,000) and Patient Safety Award building a VR platform that diagnoses Parkinson's disease through movement and voice analysis, achieving 92% clinical accuracy and capturing 3x more patient data than traditional methods.
  • Trained a TensorFlow model on motion and vocal biomarkers and built the Unity VR environment that captures patient interactions, delivering real-time diagnostic feedback with a live physician dashboard.
WALL-E
github devpost post
Python React GPT-4 API PyQT5 OpenCV C++ Embedded Systems
PatriotHacks 2024: Triple Winner (Patriot Favorite, Most Likely to be a Startup, Best Cyberpunk Theme)
  • Built a robot that detects and picks up trash, plus a file organizer that uses GPT-4 to categorize files with 95% accuracy.
  • Wrote the code connecting Python vision models to C++ microcontroller firmware for real-time navigation.
Peekabot
github devpost post
Python OpenCV AWS IoT Mediapipe C++ Raspberry Pi IoT Systems
HackOverflow 2024: Best Robot Hack Winner
  • Built a child safety robot using Raspberry Pi and Arduino that tracks toddler movements and alerts parents to dangerous situations.
  • Added pose detection with OpenCV/Mediapipe and live streaming so parents can monitor remotely via AWS IoT.
NaviguideAI
github devpost post
Java Python TensorFlow ResNet Computer Vision PyAudio ESP32
PatriotHacks 2023: Best AI-Powered Hack Winner
  • Engineered an indoor navigation system for visually impaired users using AI-powered depth sensing to map surroundings in real-time, achieving 91% spatial accuracy with 127ms response time.
  • Implemented spatial audio to deliver precise directional sound cues for hands-free navigation, validated through user testing with 15 visually impaired participants.
Optimal Path Navigation
github
Python TensorFlow Hugging Face 3D Mapping Depth Estimation A* Algorithm
  • Built an indoor navigation system that converts 2D images to 3D point clouds using Hugging Face depth estimation models.
  • Used A* pathfinding on 2D occupancy grids to plan safe routes around obstacles in under 5 seconds.
OS/161 Kernel
C MIPS Assembly
  • Implemented synchronization primitives (locks, condition variables, reader-writer locks), system call interface for user program execution, and a virtual memory subsystem with TLB management, page replacement, and swapping within a Harvard-developed instructional OS kernel.