Project Portfolio

Research Projects

• Machine Learnable Image Compression via “Alternating Transfer Learning” (accompanying paper).
  • Written for Adv ML course as part of my Masters
  • Use transfer learning to construct a deep learning architecture which can compress an image such that ML classification can be performed directly on the compressed image file.
  • Start with the paper - the code and github documentation is a work in progress.
  • Topics: Python, Pytorch, Deep Learning, Transfer Learning, CNNs, Autoencoders.

Kaggle

• Profile: Top rank overall: 759th globally (of ~180k members)

• Jane Street Market Prediction ML Competition: Silver Medal – top 1% (2021)

• Microsoft Indoor Navigation ML Competition: Silver Medal – top 3% (2021)

Machine Learning and AI from scratch

• Custom (tiny) GPT implementation (presentation of project learnings)

  • This was my project for the AI Safety Fundamentals course. I recreated an LLM from scratch in pytorch, largely following Andrei Karpathy’s tutorial.

• Race Simulation with Reinforcement Learning (video demo)

• Deep Neural Network (video demo)

• Linear Classification (video demo)

UI

• “Clue-less” Multiplayer Game (video explainer): Python, GUI, OO

• Kaggle Point Calculator (web app): R, Shiny

• Next Word Predictor (web app): R, Shiny, NLP

Other Projects

• 2048 Console Game: Python

• Caesar Cipher Decrypter: Java

• Calculator From Scratch: Excel

• Patio: Herringbone brick, 225 sqft, Stone firepit w/ built-in grill