Software Development Engineer @ Microsoft · CS grad · ex-AWS intern
Rust systems & tooling · ML from scratch · game tech · Fallout 4 reverse-engineering
I build across a few lanes — developer tooling, ML systems, game tech, and Fallout 4 reverse-engineering/modding. Most of my projects start from a practical itch: make something faster, automate a workflow, or learn a system deeply enough to bend it. Lately that means a lot of Rust and building AI/ML from scratch — writing the thing the API calls, not just calling the API.
🦀 overseerA Fallout 4 mod manager in Rust, on the Mod Organizer 2 model: non-destructive, crash-safe transactional deploys, real
|
🧠 sky-aiFrom-scratch language models: a clean-room GPT-2 (124M) that matches OpenAI bit-for-bit, plus a modern
|
🐻 KumaA lightweight C++ game engine: Vulkan rendering, an SDL3 platform layer, resource loading, input, and timing over a growing core architecture.
|
🩹 AddictolCombined Fallout 4 engine fixes — performance, stability, and crash patches — built with the Dear Modding FO4 group. My most-starred project.
|
🤖 sky-cliA Claude Code-style coding-agent CLI in Rust: an LLM agent loop with tool calling, over OpenRouter.
|
A Community Shaders port for Fallout 4 — bringing modern rendering features to the Creation engine.
|
Also worth a look — interpreter-rust (a Lox tree-walk interpreter in Rust, Crafting Interpreters) · Vault (a Next.js + Plaid personal-finance dashboard) · HouseRules (a curated F4SE/MCM config mod).
- Overseer — growing the Rust mod manager: hardening the transactional deploy engine and the
ratatuiTUI. - Sky-AI — training the from-scratch GPT-2 (124M) and scaling the custom
skyaistack toward ~1.5B. - Fallout 4 engine work — combined engine fixes via Addictol, plus performance and crash patches.
- Going deeper in Rust — interpreters, CLIs, and systems tooling.
- A terminal-first dev environment — reproducible across Windows + WSL via dotfiles.