Projects

MedSpeak

During the TreeHacks 2023 Hackathon held at Stanford, I developed MedSpeak, a real-time web application designed to help doctors sift through complex electronic health record (EHR) documents to locate the most relevant information needed during patient visits. With MedSpeak, doctors can quickly access important data without wasting time searching through EHR records, enabling them to focus more on their patients. For this project, I utilized SvelteKit for both frontend and backend development, and the OpenAI embeddings API to evaluate essential data points. Additionally, I incorporated D3.js to create visual representations of numerical data.

Cryptum

Another recent project that I worked on during the BoilerMake 2023 Hackathon is Cryptum, a file storage app that utilizes a custom hardware security module called the Tillitis TKey for secure data storage. We built the firmware for the TKey in C, and created a communication protocol for web apps using the WebUSB standard. Additionally, we used SvelteKit for the backend and frontend. Our project won the prize for Best National Security Project!

Wordle Together

I built wordletogether.com, a multiplayer version of the popular game Wordle that has been played over 200,000 times. It was built using Svelte and µWebsockets.js, and hosted on AWS. Although I had used these frameworks separately before, this project allowed me to bring them together, and I learned how to quickly go from an idea to a fully functional product.

Self Driving Car

I built a self-driving car using a Raspberry Pi 4, the Luxonis OAK-D camera, and a custom-built chassis. To collect training data, I wrote an Android application that acted as a remote control for the car over Bluetooth. I used Tensorflow to train a lane detection and following model, and ran it on the OAK-D's Myriad X VPU. I then combined these readings, allowing the car to control its throttle and steering angle. This was one of my first projects using machine learning, and I learned a lot about the process training, deploying, and optimizing models.