Hi, I'm Hal. I mostly work on data science, ML and RL with some game dev on the side. Check out some of my projects below. Feel free to reach out to me on LinkedIn or by email: hal@kolb.co.uk.

Framework inspired by the Arcade Learning Environment (ALE) that allows training reinforcement learning AI agents on Nintendo Entertainment System games, as ALE does for Atari 2600 games. Built on top of the NES emulator cynes to remove the burden of environment design when building RL agents.

Trains and evaluates machine learning models on ~500,000 Auto Trader adverts. Achieves a test accuracy of 93%.

Report analysing changes in reported crime in Leeds during the start of the COVID-19 pandemic and corresponding national lockdowns. Analysis conducted with public data from the police data archive.

• date-fuzz: Fuzzy extract dates from text corpura - Repo | PyPi
  
• fANSI-text: Provides a simple interface for formatting terminal text using ANSI escape codes - Repo | PyPi
  
• pretty-warnings: Module for displaying formatted warnings - Repo | PyPi
  

Tiny chess agent to compete in the FIDE & Google Efficient Chess AI Challenge.

From-scratch realtime GPU raytrace renderer written in Taichi Lang.

Python-based optimal route finder using OpenStreetMap road data. Written to effectively visualise and compare common search algorithms.