My personal website
Here is where I’ll show some of the projects I’ve worked on. A brief description of some of projects/repositories
Deep reinforcement learning agents applied to the task of stock-trading. I leveraged Dyna-Q to create an agent that can estimate future states with a rudimentary time-series approach, and that can use Reddit comments to regularize its decision-making. *Topics: reinforcement learning; deep learning; python; pytorch; sqlite; algorithmic trading
Uses genetic algorithms to find adversarial examples for deep neural networks in PyTorch. The core idea is also used for adversarial training. *Topics: python; pytorch; genetic algorithms; adversarial attacks; deep learning
Generates artifical python 3 documentation, using GPT-2. Built a parallelized data pipeline using Go and to scrape all of the python docs and clean the data. *Topics: golang; python; documentation; transformer methods; web scraping; data engineering
Playing around with graph neural networks. Designed a parallel algorthm to train many models in parallel and cache them for robustness analysis. *Topics: graph neural networks; python; pytorch; multiprocessing; t-sne;
Differential privacy, explored and summarized for a previous internship. This repo includes a presentation I gave to some of the portfolio companies at Georgian Partners. Learn more about them here *Topics: differential privacy; machine learning; stochastic gradient descent; research
A few of my friends went go-karting together, and we were given a paper reporting our lap times. I analyzed this data to tell a story of how we all performed. *Topics: python; data science; visualization; R; go-karts;
A comparison of two popular machine learning frameworks. So far, I’m comparing an implementation of fastText between the two. Highlights of analysis will be found here, more in-depth details will be found at the link above. *Topics: pytorch; fastText; python; keras; OOP; sentiment analysis;
An open-source contribution I’m working on. I’m aiming to optimize some numerical solvers using Cython, or numba. Topics: plasma physics; cython; numba; open-source; python; benchmarking;
A Flask app that leverages nltk’s brown corpus to illustrate simple analogical reasoning. Topics: flask; nltk; gensim; python; word embeddings; natural language processing;
A web app built with R’s shiny package to quickly plot and analyze data from fixed-width files. Topics: R; shiny; plotly; dplyr; tidyr