Projects I've worked on as a CS grad and undergrad student.
Parasite made history, but Korean dramas paved the way. A comparison of MyDramaList and Douban ratings in March 2020.
Applying the latest NLP model on the latest NLP dataset (as of spring 2019). Completed for CS224U at Stanford.
pitch deck | appendix | video demo
Spark is a smart parking system. Sensors embedded in the street, together with nearby payment kiosks, track the availability of parking spots. We provide a mobile app for drivers to find empty parking spots, make payments, get notified when time is running out, and navigate back to their parked car. For the city, the aggregated data helps parking enforcement easily identify expired parking spots and gain insight on traffic and transportation trends.
Voted "Most likely to succeed" and "Product you would be most likely to buy/use" among six ventures in CS147 at Caltech.
Rebuta is an online debate platform with a unique forum moderator: a machine learning model that determines how many upvotes a user-contributed debate argument merits.
Completed for CS145 at Caltech with three teammates; we received an A+.
Before a patient leaves a hospital, a discharge interview must be conducted. The process is lengthy, requiring a nurse or doctor to conduct the interview, record it, and later transcribe the audio to text by hand. My team created a tool that transcribes the conversation in real time and at the end of the interview, immediately produces a formatted document. I developed the back-end algorithm that processes raw text, matches the conversation against a list of standard discharge questions, populates a form with the patients' responses, and generates suggestions for post-hospital care.
Won the Amazon Alexa Best Voice User Experience award and honorable mention (aka lightsabers) from the IBM Watson judges at TreeHacks 2016.
I designed and implemented a library that decomposes a protein prediction algorithm into bite-sized, parallelizable tasks, distributes them to different machines, and reassembles intermediate and final output files. TARDIS is short for "TAsk Re-DIStribution."
Named Sidney and Nancy Petersen SURF Fellow.
Popularity of a food, defined by how often it is mentioned in Yelp reviews, is charted over the last five years. From there, the user can dive in to explore how various restaurants and shops dominate the market.