Chudi Nnorukam

school Background

I studied Data Science at UC Berkeley, where I developed a foundation in statistical modeling, machine learning, and data engineering. After graduating, I moved into production software — building systems that ship, not just notebooks that demo.

My path has been nonlinear in the way ADHD paths tend to be: I've built algorithmic trading systems that manage real capital, SaaS products that process bank statements, bug bounty automation frameworks, and the site you're reading now. The through-line is building systems where correctness matters — where a wrong output means lost money, missed vulnerabilities, or broken trust.

I have ADHD, and I write about it openly because the intersection of neurodivergent cognition and AI tooling is where I do my most original thinking. The pattern recognition and systems thinking that come with ADHD are genuine advantages in AI-assisted development — if you build the right scaffolding around them.

deployed_code What I Build

My work falls into three categories:

• 01 AI-visible web systems. chudi.dev is the flagship — a site built for human readers, LLM retrieval, and AI agent interoperability. It implements llms.txt, structured data for AI citation, WebMCP browser tools, and answer-first content design. I document the architecture publicly so others can adopt the patterns.
• 02 Production trading systems. Polyphemus (Polymarket) and a CME Micro Bitcoin Futures bot — both handling real capital with signal detection, risk controls, and automated execution. These taught me that software correctness isn't academic when your balance is on the line.
• 03 Developer tools and SaaS. StatementSync (bank PDF to spreadsheet), citability.dev (AI visibility auditing), BugBountyBot (semi-autonomous vulnerability research), and MicroSaaSBot (multi-phase product development). Each is a production system with users, not a side project with a README.

engineering How I Work

I use Claude Code as my primary development environment — not as a novelty, but as a production workflow. I've written about how this reduces error rates by 84% compared to manual development in my workflow, using quality gates, context management, and multi-agent orchestration.

My stack is SvelteKit, TypeScript, and Tailwind for the web; Python for trading systems and automation; Supabase and SQLite for data; and Vercel for deployment. I choose tools that let me ship fast without accumulating debt that slows me down later.

I'm available for consulting on AI-visible web architecture, Claude Code workflow design, and custom agent-facing system builds. I work with founders, developers, and small teams who want technically credible systems that are legible to both humans and machines.

lightbulb Why chudi.dev Exists

Most personal sites are either portfolios or blogs. chudi.dev is an architecture case study. Every page is designed to be simultaneously readable by humans, extractable by LLMs, and queryable by AI agents. The machine-readable surfaces (llms.txt, structured data, WebMCP tools) aren't decorative — they're tested daily by an automated audit pipeline.

I publish the architecture documentation in a separate open-source repository so anyone can inspect the claims, run the verification checks, and adapt the patterns. The goal isn't to be the only AI-visible site — it's to document the approach well enough that it becomes a replicable standard.