Projects
Deep Learning for Congenital Heart Disease
Project goal: build a deep learning model to classify ultrasound videos as one of five different pathologies
Meme Image Caption Generator
Project goal: take a base image template and create a model to automatically generate a funny caption
Spotify Track Recommendation
The goal of this project was to build a recommendation system that would provide the most relevant set of songs given a playlist. Completed as a part of Harvard’s Data Science 1 course, AC 109a, our team explored collaborative filtering, content-based filtering, and k-means clustering to create our recommendation system.
Reinforcement Learning Implicit Stock Trends
The goal of this project was to train an agent using reinforcement learning to identify and capitalize on implicit trends in stock histories. Completed as a part of Havard's Artificial Intelligence course, CS 182, our team defined our own Markov Decision Process and tried solving it with Q-learning and Approximate Q-learning.
Amazon Product Rating & Review Statistics
The goal of this project was to investigate the psychology behind Amazon product ratings and reviews. Completed as a part of Harvard's Linear Models course, STAT 139, our team carried out a fleet of statistical tests. Most notably, we performed a sentiment analysis of review text and incorporated that into a multiple linear regression, together with other engineered features to identify the correlates of positive product ratings.
Python Library Development: Automatic Differentiation
The goal of this project was to develop a python library that could perform automatic differentiation - a method for taking derivatives quickly and to machine precision. Completed as a part of Harvard's Systems Development for Computational Science course, CS 207, our team built a user friendly package that solves an explicit computational graph to propagate derivatives.