What I am doing now

Building Jido

My main open-source focus is Jido, an agent framework for Elixir and the BEAM. I am working on the parts of agent systems that become important after a demo works: coordination, supervision, durable execution, observability, and recovery from failure.

I believe the hard problem is not making one agent look smart. The hard problem is helping many agents and people work together without producing more noise. Jido is my way to explore that problem in public and turn the useful answers into software.

Current work includes:

• Making the core agent abstractions smaller and easier to understand.
• Improving the tools for long-running, event-driven, and multi-agent workflows.
• Building practical examples that show why OTP and supervision are a strong base for agent systems.
• Helping contributors use Jido in real applications, not only experiments.

Leading an AI-first engineering team

I am a Director of Engineering at ServiceCore. My work there is about building a strong senior team, creating clear operating systems, and using AI where it improves customer and business outcomes.

I am interested in the organizational side of AI as much as the technical side. A company does not become AI-first when it buys more tools. It becomes AI-first when people can identify good problems, redesign the work, measure the result, and learn quickly.

Writing what I learn

I am writing about AI agents, Elixir, engineering leadership, and the systems needed for people and agents to work together. I want the writing to come from work that I have done, including the parts that failed or changed my mind.

I am also updating this site so it better represents my current work. That includes a public statement on how I lead and more useful technical notes.

Questions I am exploring

These questions have most of my attention:

• What should an agent runtime own, and what should stay in application code?
• How do teams keep human judgment and accountability when agents do more of the work?
• What can agent builders learn from decades of actor-model, workflow, and distributed-systems research?
• How do we measure whether AI reduced work instead of moving it somewhere less visible?

Current defaults

I am protecting more time for deep work, open-source maintenance, writing, and family. I am saying no to work that does not support those priorities. I am still glad to meet builders and engineering leaders who are working on durable AI systems.

If that describes you, send me a note.