I build local-first AI products with explicit safety boundaries, observable workflows, and practical developer tooling.
I am an independent systems architect focused on turning agentic ideas into software people can inspect, run, and control. My work prioritizes local ownership of data, narrow permissions, repeatable validation, and designs that remain useful even when a model is uncertain.
| Project | Focus | What it demonstrates |
|---|---|---|
| Kryptic | A local-first AI engineering agent CLI | Bounded execution, evidence-based verification, protected workspaces, and provider separation |
| Kairos | A local AI-agent control plane | Skills-aware workflows, local data tools, configurable providers, and visible approval gates |
| Aion | A compact coding language model project | End-to-end model training and practical ML systems thinking |
| MCP Config Auditor | A local CLI for reviewing MCP configurations | Deterministic security checks for credentials, shell evaluation, remote transport, and permission bypasses |
| MCP Change Review | A local CLI for reviewing MCP configuration updates | Secret-safe change reports for new servers, command changes, permission expansions, and network-target changes |
I believe good AI tooling should make the operator more capable without making them less informed. I design for local control, explicit approvals, clear failure states, and reproducible checks rather than opaque “autonomy.”
I am interested in small, high-signal collaborations around developer tools, local AI, trustworthy agent workflows, and privacy-preserving productivity software. If you are evaluating an idea or need an engineering partner for a focused prototype, open an issue on a relevant repository with context and a concrete goal.
JavaScript · TypeScript · Python · Node.js · Docker · Local-first architecture · AI agents · MCP · Safety engineering · CLI tooling