Projects
Staying hands-on means continuing to learn, build, and improve. These projects are how I keep my technical judgment current and stay better connected with my teams.
CiteLoom
Section titled “CiteLoom”I built CiteLoom to make AI-assisted research easier to trust. It gives people a private place to work with their own documents while keeping every answer connected to evidence they can inspect.
RepoChord
Section titled “RepoChord”RepoChord grew out of the coordination problem that appears when one change crosses several repositories. It divides the work among isolated repository agents, brings their results back for review, and keeps integration under human control.
CrewAI Development
Section titled “CrewAI Development”I created CrewAI Development to remove the setup work that gets in the way of experimenting with agent systems. It packages CrewAI, its development tools, and a ready-to-use editor into Docker, treating the container as an isolation boundary rather than only a deployment format. Agent execution stays inside the environment, while access to host files and services is controlled through explicit mounts and network configuration.
View CrewAI Development on GitHub
KodaCode
Section titled “KodaCode”KodaCode is where I explore what a coding agent looks like when trust is built into the runtime instead of left to a prompt. It makes permissions, tool use, orchestration, and session history visible so developers can understand how the work was done, not only what changed.