Inspectability
A useful system should show its reasoning, state, and failure modes.
About
I build open tools for agents, data, and evidence—and I care whether the decisions they support can be inspected, reproduced, and challenged.
I’m a software engineer at IBM on the Developer and Operator Experience team, where I’m a core maintainer of Langflow. My work sits at the boundary between the clean API a developer wants and the messy production system that has to keep working: agent tools, MCP, authentication, model providers, retrieval, packaging, SDKs, and release pipelines.
Before agents, I worked on statistical evidence. I earned dual undergraduate degrees in Statistics and Computer Engineering from the University of Washington, then a PhD in Statistics and Computer Science at Iowa State. My dissertation developed reproducible methods for comparing 3D scans of spent bullets. That research now has a second life in Verity, an open engine for calibrated forensic surface comparison.
I also build Aeroza, a weather platform organized around a similar idea: forecasts should be queryable, testable, and scored against what actually happened. The domains are different; the instinct is the same.
I live north of Seattle with my partner and our two cats. Away from the terminal, I’m usually gaming, following Pacific Northwest weather, or convincing myself that one more side project is a reasonable way to spend a weekend.
Principles
A useful system should show its reasoning, state, and failure modes.
Measure uncertainty and publish the misses, not only the attractive examples.
Good research and good developer experience both improve when the code can be run.
MCP, orchestration, policy, and the infrastructure underneath a reliable tool call.
Interfaces that help people reason about variation rather than hide it.
The best technical explanation leaves someone able to do the next step alone.
Timeline
2025 — Present
Core Langflow maintainer on Developer and Operator Experience. I work across MCP, the Langflow SDK, agent tool-use policy, model providers, retrieval tooling, and the production seams users feel first.
2022 — 2025
Built Astra DB and GenAI product surfaces, then joined the Langflow core team after DataStax acquired the project in 2024. Moved with DataStax into IBM in 2025.
2021 — 2022
Built serverless Python functions for scientific computing and wrote much of the platform’s engineering story.
2017 — 2021
Led statistical consulting, Shiny application work, and technical teaching. Helped seed the open-source OmniacsDAO suite.
2012 — 2017
Developed reproducible methods for forensic bullet comparison with CSAFE and shipped the work as papers, R packages, and a dissertation.