🧑‍💻 About Me

Prospectus

HoyaHacks

Prospectus was my very first hackathon project, where I jumped right into building an ambitious AI-powered college admissions assistant. Our idea was to leverage React, Django, Azure OpenAI, and custom vector embeddings for smart, query-based guidance about universities.

While we didn't end up with a fully functional product, the experience was incredibly valuable. I learned a ton about designing architectures, working with unfamiliar tools, and collaborating under real-world constraints. Even though we only got a frontend demo and a high-level architecture, it gave me first-hand insight into how complex building real applications can be.

Prospectus motivated me to level up my technical and teamwork skills for future hackathons. That first weekend set the stage for actually shipping working projects at later events.

Tech Stack (planned and prototyped):

SweetFriend

PennApps

SweetFriend is a full-stack diabetes management app that gives users real-time glucose tracking, AI meal recognition, and instant health insights. I led the frontend development using React, building a responsive dashboard, custom chart components for live blood sugar graphs, and an intuitive photo upload feature for meal recognition powered by Tune Studio's vision model.

I implemented the integration with the Dexcom API, bringing continuous glucose data into the app and linking it with event logging. The Flask backend connects user events and meal predictions, stores data in MongoDB, and uses Cerebras for ultra-fast AI health advice. I also helped wire up Twilio for notifications, alerting users about critical changes in their glucose levels. Shipping a polished MVP with this level of integration in just 36 hours took grit, and the project made me much more confident with complex health tech stacks and real-time API workflows.

Tech Stack:

Albatross

HackHarvard

Albatross is an AI-powered navigation app that helps users avoid high-risk areas in any city by finding the safest routes using real-time crime data. My role was building the Vue.js frontend, where I designed and implemented the interactive dashboard that overlays crime polygons and safe routes on a dynamic map. I connected the frontend with a backend pipeline using AWS and Databricks for data processing, MLflow for clustering high-crime zones, and Cloudflare Workers to serve the routing data to the HERE Routing API.

I focused on getting geospatial data and route overlays working smoothly so users could instantly see where to travel more safely. We built and deployed a working demo for New York City during the hackathon, and even in demo mode the system visualizes hot zones and safer paths in real time. Building Albatross pushed me to the next level with map rendering, geospatial data, and full-stack cloud integration.

Tech Stack: