Thesis with Prof. David C. Parkes Research Group
AI Pricing Collusion: Multi-Agent Reinforcement Learning Algorithms in Bertrand Competition

• Read Thesis

• Wrote thesis on reinforcement learning multi-agent collusion in the Bertrand competition economic environment.
• Extended research included in IJCAI, AAAI, and NeurIPS conferences (paper, OpenReview).
• Worked to uncover heightened collusion using deep reinforcement learning algorithms in discrete and continuous settings.
• Mitigated collusion through supervisor mechanism and other environmental mechanisms.
• Earned highest honors in Computer Science concentration

CS182 Artificial Intelligence Project
Intrinsic Curiosity Module with Proximal Policy Optimization (PPO)

• Read Paper

• Substituted extrinsic rewards with "intrinsic" reward based on prediction error of learning agent, using PPO RL algorithm.
• Applied this to OpenAI Gym's MountainCarContinuous game and saw significant improvement over vanilla PPO, since the game definition encourages exploration.

CS109a Data Science Project
Predicting Crime

• Visit Website

• Used crime incident reports, weather data, and hospital, police, and school locational data to predict type and frequency of crime.
• Trained a random forest model to reach results.

CS136 Economics and Computation Project
Recommender System for Harvard Courses

• Read Paper

• Used the Harvard Q Guide with Collaborative Filtering and used Previous Coursework with Content-Based Recommendation to provide ranked course recommendations for students.
• Created dummy dataset to work with, recommender took into account concentration requirements.
• Obtained accurate results, hope to use it in the Course Search tool at Harvard.

CS171 Visualizations Project
The Rise of Tech Giants Through the Lens of M&A

• Visit Website

• Used historical M&A data, stock price, market trends, industry trends, etc. to tell the story of the rise of big tech.
• Created website using D3.js library, with interactive visualizations.

Harvard Data Ventures Project
Harvard Square Homeless Shelter

• Used homeless shelter anonymized attendance data to gain insights on trends and to improve the lottery system.
• Made significant impact on the quality of the shelter's service.
• Continuing to find ways to improve the homeless shelter through data collection and analysis.

Applied Mathematics 50 Project
Applying Betting Theory to March Madness

• Read Paper

• Applied various betting techniques to determine the best strategy on betting in the NCAA Tournament.
• Used MATLAB to simulate the strategies on 2018 March Madness data.
• Found that Kelly Criterion betting, based on Shannon Information Theory, is most effective.

CS50 Project
Progree - Progress to Graduation

• Watch Demo

• Created an interactive website to track progress to college graduation.
• Displays information such as percentage of time spent in each department, average workload, rating, difficulty, etc.
• Currently only works for Harvard classes, but can easily be implemented with other universities.

Stories from the End of the World Project
A Short Story, The Surface

• Read Story

• Short story about a world underwater.
• Won award for excellence in the course.

Harvard Sailing Club

• Founded and grew sailing club for beginners or casual sailors.
• Managed grants and outings for the club.
• Personally sailed since 13 years old, have sailed the Long Island Sound and Fire Island