I build practical software systems at the intersection of AI agents, developer tooling, automation, and product engineering.

My current public work is centered on Andromeda, a local-first agent runtime and CLI written in Go. It explores tool execution, provider routing, MCP integrations, plugin and skill systems, memory, workflow orchestration, sandboxing, telemetry, and a terminal UI built for real development workflows.

I care about software that is useful under pressure: clear command surfaces, strict permissions, reproducible workflows, readable architecture, and systems that stay understandable as they grow.

Most of my product experiments and operating systems live in private repositories while they mature. Public work here is intentionally curated around reusable foundations, infrastructure, and agent tooling.

• Building Andromeda as a robust agentic development runtime.
• Designing local-first workflows for agents, tools, memory, permissions, and automation.
• Turning messy product ideas into small, testable systems with clean interfaces.
• Using AI as an engineering multiplier while keeping control, traceability, and reviewability in the loop.

A Go-based CLI and runtime for agentic development workflows.

• Agent runtime, workflow execution, memory, indexing, and context tooling.
• Provider abstraction for cloud and local model backends.
• MCP client and tool bridging.
• Plugin, skill, package, permission, secret, sandbox, and telemetry layers.
• Bubble Tea terminal UI and signed release pipeline.

• Start with a narrow, useful workflow.
• Make interfaces explicit before scaling implementation.
• Keep permissions, secrets, and operational risk visible.
• Prefer tests, traceability, and readable architecture over clever shortcuts.
• Ship small increments that can survive real feedback.