A lab of one.

I'm a software engineer working independently on AI infrastructure, engineering tooling, and dataset systems. EffNine is my workshop — not a company, not a product, just a place where I build things I find interesting and useful.

I tend toward systems that stay out of the way: tools that do one thing well, infrastructure that's quiet and reliable, and research that's honest about what it knows and doesn't know.

My work sits at the intersection of AI and software engineering tooling. I build systems that help developers — and AI agents — work with code more effectively: engineering context layers, personal memory infrastructure, evaluation frameworks, and versioned dataset pipelines.

I also work on model training and evaluation. Atlas started as a personal frustration — every time I wanted to experiment with a new model, the dataset was locked to one format, undocumented in its processing, or impossible to reproduce — and grew into a governed dataset system with 9.5M+ records.

Right now I'm focused on three things: CodeBro (engineering context and memory for AI coding agents), Conductor (an OpenAI-compatible AI gateway with multi-provider routing), and MyContext (personal context infrastructure for AI).

Atlas continues in the background — the dataset is released and the pipeline works, but full model training is paused until the evaluation story solidifies.

Rust, Go, Python, TypeScript. MCP for agent tooling. Cloudflare Workers and Pages for deployment. SQLite and JSON for storage (when it fits the problem). No framework dependencies unless they solve something real.