Projects

Extended Yale Asynchronous VLSI (AVLSI) group's physical database representation of circuit designs to include geometric information. Converted routing information into set of non-overlapping rectangles, modeled wires as resistances with nearby wire segments resulting in coupling capacitances. Implemented in C++.

Built genetic algorithm to optimize traffic signal timing in a road network. Used Simulator for Urban Mobility (SUMO) open source project to simulate traffic flow through Yale's campus, optimized various aspects of signal timing. Achieved high performance relative to benchmarks. Implemented in Python with pyeasyga to facilitate genetic algorithm dynamics.

Designed graph neural network (GNN) architecture for classification of activity from brain scan. Used image processing pipeline of previous work to generate graphs from fMRI data to train/evaluate network. Outperformed many other GNNs tested by pipeine designer including Chebyshev convolutional network, graph attention network, and GraphSAGE. Implemented in PyTorch-Geometric (PyG).

Built Chrome extension that adds recommended courses into the Yale Course search interface. Uses cosine similarity between course title + description embeddings. Embeddings generated with tensorflow universal sentence encoder. Extension (implemented in JS) queries recommendation server implemented in Flask/Python to retrieve recommendations.

Designed and built (in group) infrared-ultrasonic hybrid system to facilitate indoor localization and navigation. Infrared used for communication with beacons while ultrasonic used for time of flight distance sensing. Augmented sensor system with computer vision. Raspberry Pi used for onboard control, arduino for individual sensor systems. Majority of software in Python, some C.

Designed and built (in group) infrared-ultrasonic hybrid system to facilitate indoor localization and navigation. Infrared used for communication with beacons while ultrasonic used for time of flight distance sensing. Augmented sensor system with computer vision. Raspberry Pi used for onboard control, arduino for individual sensor systems. Majority of software in Python, some C.

Implemented scripts with Python/qiskit to probe error rates and scheduler properties of IBM quantum machines. Motivated by investigating possibility of device fingerprinting using geographically specific cosmic ray strike events.

Built RFID based door lock using Arduino UNO, RFID reader/tags, servo, magnetic sensor, 3D printed lock and wooden demo door. Used UID of RFID tag to identify user and make decision about whether to unlock or lock based on ID. LED and buzzer also used to indicate status of lock. Servo controls physical lock, awaits closed signal from sensor to re-lock following an unlock.

Built relative entropy analysis program for Royal Society corpus of XML annotated scientific texts. Mapped the cognitive surprise of 30 scientific disciplines over >250 years. Built in Python using matplotlib, beautifulsoup, and scipy.

Built webpage that scrapes the Yale Daily News homepage for article titles, and bot-generates fake articles using the real titles as a seed. Real names are removed for privacy, and hyperlinks all go back to the homepage. Built in Flask and hosted on AWS.

Developed a full stack (Python/flask) web (HTML/CSS/JS/Bootstrap) and iOS (Swift) streaming service for instructional baseball content from professional/collegiate level coaches (no longer live).

Founder and CTO, led development team on full stack video chat web platform to host social networking events for college students. Built on React.js and Python serverless backend powered by AWS (not live).

Built full stack web (HTML/CSS/JS/Bootstrap) application on flask/Python to connect Yale students with local high schoolers in need of tutoring (no longer live).

Developed serverless API on Python powered by AWS for mobile wellness app built on React Native (not yet live).

Built mathematical rendering of ideal scrambler amusement park ride in Desmos, as well as Fourier/polynomial curve-fitting analysis of real kinematic data collected from ride in Python.