Now
Updated June 2026 · Boulder, CO
Research
EdgePlane
A production-grade control plane for multi-agent fleets. The driving question is what it takes to run agents with real operational rigor: the layer between a demo that works once and a system you can trust in production. Agents are first-class principals with stable identity, an inbox, and durable task state.
EdgePlane · how the fleet runs
- Agent evaluation. Developing methods for measuring agentic system reliability in enterprise contexts, where failure modes are subtle and the cost of errors is real. Most existing evals assume clean inputs and well-specified tasks. Production rarely looks like that.
Forward Deployed Work
- Embedded with enterprise clients building production AI applications, shipping working systems rather than decks. Current engagements involve agent-native workflows in organizations navigating complex legacy infrastructure and real data governance constraints, codifying what works into repeatable deployment patterns. If you are building enterprise AI systems and want a practitioner in the room, reach out.
Thinking About
- What the right abstraction layer for enterprise agent orchestration looks like, and why the gap between "it works in a notebook" and "it runs in production" is still so wide for most organizations deploying LLMs at scale.
- How frontier model capability improvements are reshaping the economics and architecture of agentic systems. The answer is shifting faster than most enterprise deployment patterns can keep up with.
- The organizational conditions that determine whether enterprise AI deployments compound or decay. High agency and the ability to navigate ambiguity in complex organizations matter as much as the technical implementation.
Writing
- Publishing regularly on this blog. The focus is what I am actually building and deploying: agent infrastructure, secrets management, orchestration patterns, and the real failure modes. Field reports, not tutorials.
- Also active on @rmerlin on Twitter/X for shorter takes and the AI builder conversation.
Inspired by nownownow.com.