A wrist-worn electrical impedance tomography system that uses a four-electrode array to capture forearm muscle activity and classify micro-gestures in real time. The project spans custom firmware, synchronized data acquisition, signal preprocessing, and a machine-learning pipeline for reliable gesture recognition.
An autonomous drone built on a custom SpeedyBee F405–based flight stack with ArduPilot and a Raspberry Pi vision module. The system uses onboard computer vision for real-time perception and sends MAVLink control commands for fully autonomous navigation and gesture-based interaction.
Carte Diem is a add-on module that integrates smart features to regular carts. This device eliminates the need for checkout lines while maintaining anti-theft features. It’s modularity allows stores to easily implement this technology into the existing ecosystems.
Redesigning the market-leading Sartorius AG Incucyte into a next-generation, miniaturized live-cell research platform. The project develops a custom embedded control system with precise thermal, humidity, and airflow regulation to enable reliable long-term cell imaging in a compact form factor, with technical details shared only within NDA constraints.
Turn any flat surface into an interactive touchscreen by combining LiDAR-based position tracking with vibration sensors that classify taps, drags, and other gestures in real time.
Our computer vision research project developed in Python enhances real-time hand tracking and gesture classification using MediaPipe, YOLOv8, and DeepLabCut to support posture correction and performance analysis for musicians. Project Research paper accepted to HCII 2026 conference for publication.
An advanced content-aware, image resizing program utilizing seam carving algorithm in C++ that allows image to be scaled without losing/distorting meaningful image landmarks.
This project is a natural language processor implemented in C++ for predictive classification using the naive bayes algorithm. The classifier has been tested on large .csv datasets and has demonstrated an accuracy of 87% with O(n) processing time.
Built a RESTful web server and custom doubly linked list in C++ to manage an interactive office hours queue, focusing on dynamic memory, iterators, and real-time client-server communication.
