Writing

This is the collection of assorted articles that I have written, dating back from early graduate school projects to more recent investigations. Some of these articles are self-contained write-ups of one-off explorations, while others document a single stage in a larger scale project. Articles are organized chronologically, but check out Projects if you want to view articles grouped by project.

2026

3D Pose Harness

3D Pose Harness

Creating a 3D pose harness for robust estimation.

Scaling Local Compute

Scaling Local Compute

Effectively utilizing a second machine.

Implementing Flash Attention

Implementing Flash Attention

Adding flash attention to our LLM.

Local ARC Challenge

Local ARC Challenge

Tackling the ARC 2 benchmark with local compute.

Building Tokenizer from Scratch

Building Tokenizer from Scratch

Implementing a tokenizer for my LLM.

2025

Building Large Language Models

Building Large Language Models

Implementing LLMs using Raschka’s book combined with my own developer preferences.

SAM 3D

SAM 3D

Evaluating a new state-of-the-art model for human 3D pose estimation for lifting form analysis.

Nanochat

Nanochat

I trained my own LLMs from Karpathy’s repo.

Perfect Form: To the Cloud

Perfect Form: To the Cloud

Upgrading from local prototype to cloud deployed application.

2024

Advent of Code

Advent of Code

Participating in advent of code for the first time.

AWS Community Day

AWS Community Day

I presented at AWS Community Day.

AWS Certifications

AWS Certifications

I obtained a few AWS certifications.

2023

Perfect Form: Computer Vision for Powerlifting

Perfect Form: Computer Vision for Powerlifting

Building a useful tool for analyzing powerlifting biomechanics.

2021

Reinforcement Learning Notes

Reinforcement Learning Notes

Reinforcement notes from Pieter Abbeel’s lectures.

2020

Lost Cities Part 2: Learning from Deep Q-Networks

Lost Cities Part 2: Learning from Deep Q-Networks

Adding eyes to our deep Q-network for a physical face off.

Lost Cities: Using Deep Q-Networks to Learn a Simple Card Game

Lost Cities: Using Deep Q-Networks to Learn a Simple Card Game

Using reinforcement learning to solve an imperfect information card game.

Steep Part 2: Imitation Learning to Complete Rocket Suit Races

Steep Part 2: Imitation Learning to Complete Rocket Suit Races

Controlling a rocket suit through virtual courses with neural networks.

Steep: Imitation Learning to Complete Ski Races

Steep: Imitation Learning to Complete Ski Races

Training neural networks to (virtually) ski.