• 👋 Hi, I’m @torial
• 👀 I’m interested in ML, algorithms, Python, Rust, V, C#, Godot
• 🌱 I’m currently learning Rust, V, ML, and Godot
• 💞️ I’m not currently looking to collaborate
• 📫 At gmail.com.

I often like solving hard problems (hence my interest in ML and algorithms). I also enjoy refactoring existing code bases, optimizing performance, and starting projects from scratch. Many times in investigating complex performance issues in .Net and python, reducing the memory pressure for the garbage collector was a successful strategy -- this led me to look into alternative memory management strategies like that used by Rust and V.

1. iui some incremental fixes for updated V version.
2. lazypredict initially just fixes for a small project I was doing for a client. Additional changes forked it to support more things like hyperparameter defaults. Needs more refactoring to be fully generalizable. At this point more of a fork than a patch.
3. cobra-language initial attempt at upgrading the installer to work with newer .Net versions.

I am a big believer in learning by building technology from scratch. I've built things like messaging system, template libraries (handlebars compatible), translation leveraging systems, web server, one time password and associated handshaking, Q Learning (Reinforcement Learning), neural networks, GPT 2, Decision Trees, and simple XG Boost. Such an approach increases my understanding of underlying technologies - some of these are used in production projects -- the others are listed here.

1. liteml - initially implemented following a video from Andrej Karparthy, improving performance, and adding decision trees and a simple XGBoost implementation.
2. liteGPT - initially implemented following a video by Andrej Karpathy implementing GPT-2, additional tweaks including GPU support.
3. cobraide - an initial IDE implementation for the defunct Cobra programming language, basic support for autocomplete, UI forms design, code folding, project folder, and task list.