Projects

Predicting NYC Cab Fares

Implemented 4 versions of a basic Neural Network using Tensorflow to predict cab fares for every time bracket of the day with 4 different data cleaning methods, resulting in the best output of a Root Mean Squared Error of 2.7 among the 4, placing it in the top 5 Kaggle submissions among almost 1500 submissions.

Modeling COVID-19 Spread at UCLA

Developed a network of student interactions in classrooms using stochastic block models and the NetworkX library in Python, and implemented an SIR model in Javascript using networks in D3 to animate 10 weeks of coronavirus spread.

Optimization Library

Developed a simple optimization library in Python to implement several optimization algorithms like gradient descent, Newton’s method, and conjugate-gradient methods.

Analysis of Presidential Candidates Using Their Twitter

Scraped twitter feeds of all US Presidential Candidates, and generated Personality Insights using IBM Watson's Personality Insights API. Also developed interactive bubble and bar charts using ChartJS, along with front-end component to display information about personality traits and keywords most frequently used by candidates

Hidden Student Costs

Estimated different costs of being a student based on what state or country you're from, and your living preferences. Created interactive map using Leaflet to show distribution of college expenses, a customized cost calculator in Javascript and interactive bar charts using ChartJS.

Squad

Winner of Spectra Hackathon 2019, Best Use of HERE API, Squad is a web app that captures your location to connect you with registered callers nearby in case of an emergency or for support if you are walking home alone, or it lets you call the closest police station.

ElmoBot

Developed a chatbot that matches you to the movie character you are most like using IBM Watson’s Assistant API, Personality Insights API and IBM Cloudant in Python - Scraped thousands of movie scripts, separated each character’s dialogue and fed it to the Personality Insights API. Similarly, generated Personality Insights using user conversation with the chatbot - Wrote a least squares algorithm to match the movie personality user is most similar to

Focusbunny

Developed the website using HTML, CSS, JavaScript and Bootstrap and integrated a Stripe payment portal (site is live at focusbunny.com) for a startup (fidget spinner with a pen for better focus) - Developed a logo, design and illustrations for the page, created the site map and a wireframe and optimized search engine results.