At Penn, we have a big course directory that is sometimes hard to search through, especially given the complex nature of planning courses for the semester. Penn Courses LLM seeks to simplify this complexity by offering a chat interface to answer any course-related questions a student might have.

The US experienced $100B annual loss to Medicare and MedicaidFraud. At the same time, the process for detecing such fraud uses either manual or rule-based systems. Robin seeks to introduce ML to healthcare fraud detection, to catch fraud within minutes instead of days. My main role at Robin is to build out our product experience, including the dashboard UI panel and APIs for dataset processing and model inference.

H1B Sponsorship Data is quite few and far between across the web and there are not a lot of applications that help visualize a birds-eye view of the H1B landscape in the USA. With over 582,000 individuals living in the US on a H-1B visa and 85,000 new visas being issued per year, there is a significant need for reliable and consolidated resources to help these individuals find their next home. As part of CIS5500: Database, we created an application that can visualize a bulk of H1B Visa related statistics (over 5 million data points).

Wheelchairs can be hard/clunky for disabled people to operate. As part of ESE3600: TinyML, we ntroduce a prototype of a bot that can move based on gestures to lift this burden completely. The user will be given a remote and depending on the path the user draws using the remote, gesture bot can follow the direction of the path

This is a fun project I did to help out a high schooler with her research project into sustainbility stores. We collected data from a few directory websites of stores with their sustainable ratings, aggregate them into a single database and visualize them across the map. I had the chance to experiement with a serverless architecture (Prisma) with this app!

AVP, a victim service agency based in West Philadelphia, was seeking a solution to simplify their referrals tracking system that was spreadout over multiple Google Sheets and Forms. As part of Hack4Impact, I was leading a team to build out a solution to simplify the referrals system into 1 unified platform. Here, I built some of the dashboard UI components and APIs for tracking referrals updates. I also spent a lot of time talking with our clients at AVP and getting to know the nonprofit landscape in Philly!

OHQ revolutionizes office hours by replacing physical lines and sign-up sheets with a virtual queue system. It lets students join with a query and estimates wait times, alerting them via SMS when their turn is up, along with updates and staff announcements. For staff, it provides queue management, performance tracking, and review tools. As a backend team member at OHQ, I had a blast working with Python and Django, developing APIs for analytics, TA reviews, and enhancing existing ones, scaling OHQ to help students and university admin have a better time with OH!

I joined The DP (Daily Pennsylvanian) as director of engineering, responsible for developing/maintaining technology infrastructure. Over my time, I was able to develop a few services for the DP such as the DP+ app (contributions to UI as well as a GraphQL API to query articles from the website into the app), Narrative AI integration (making AI-generated audio available for all articles), dropcap (a service track the top 10 most viewed articles within a month) and content-rec (a service used to recommend relevant articles based on vector similarity search)

This project is a visualization of the Penn Courses Directory, connected by their relationship in whether they are prereqs to each other or whether their content is similar. The relationships are then processe into a graph-like format and visualized via a web UI with filter options, entirely with vis-react.

InstaLite is an Instagram clone, backed by functionalities like social graph management, chat, and dynamic post ranking, post management, hashtagging, commenting and liking.