AI systems & infrastructure • Distributed systems • Performance • Developer tooling
Software engineer focused on AI systems and infrastructure. I enjoy building systems that are reliable, high performance, and designed to hold up under real use.
My work sits around distributed systems, performance, and developer tooling, with a strong interest in how good architecture compounds over time. I like working close to the systems layer and thinking deeply about scalability and long-term design.
I build tools because I need them. When I hit the same friction twice, I turn it into a library, framework, or SDK that makes the next project faster and more predictable. I build and contribute to open source, and I spend a lot of time turning ideas into working systems—especially around agents and developer infrastructure.
|
AFK (Agent Forge Kit)
Production-grade framework for building deterministic, observable agent systems in Python. Designed for orchestration beyond a chat loop. 
|
socioy/joor
High-performance, type-safe backend framework for TypeScript/JavaScript designed to run across multiple runtimes without giving up speed or correctness.
|
|
socioy/bumblebee
AI-powered Python package for natural, human-like mouse and keyboard control. Built for agentic automation where "robotic" UI scripts break.
|
cui
Cross-platform React UI framework that keeps web and native components aligned behind a single API. Focuses on consistency and shared primitives.
|
|
docsy
Documentation site generator built on Astro from MDX. A Mintlify-compatible alternative with a straightforward, repo-friendly workflow.
|
tcon
Zero-dependency Rust configuration compiler built for predictable outputs. Useful for config validation without pulling in a heavy runtime.
|
- Determinism & Correctness: Systems must do what they say they do.
- Observability: If it breaks, it should explain why.
- Performance: Speed is a feature, minimal overhead is a requirement.
- Ergonomics: Good developer experience shouldn't compromise the safety of the underlying architecture.
If you want to collaborate, ship something ambitious, or exchange ideas, feel free to reach out. The fastest path is usually a small issue describing the goal and constraints, or a focused PR with a short rationale.
The problem defines the tool, not past experience. I try to stay flexible, keep solutions simple, and optimize for systems that are easy to run and hard to break.