Enterprise AI Engineering
I write and speak about how enterprise AI gets built.
The work starts below the application surface. I use this model to explain the infrastructure, runtimes, and control layers that make AI usable in production.
Ideas
Start here.
Begin with the core framework, then explore recent essays, build notes, and security briefs on the systems that make enterprise AI work in production.
Why MCP Tool Access Does Not Replace Runtime Identity
MCP can narrow which tools an agent may call, but it does not replace runtime identity, delegated user access, or downstream system permissions.
Read the pieceRecent writing
Recent essays, frameworks, and operating notes from the archive.
Agent Identity Patterns: Which One to Use, and When
The real choice is not whether an agent has credentials. It is which identity pattern fits the ownership boundary around the action.
The AI-Ready Enterprise Stack
Most enterprise AI conversations start at the model layer. They should start several layers lower.
What Every Agent Runtime Should Share
Triggers, validation, recovery, and auditability belong in the scaffold, not re-invented agent by agent.
Why MCP Is a Trust-Boundary Problem
The problem is not connector count. It is identity, scope, trust boundaries, and what the runtime is allowed to do.
Watch
Video briefings and walkthroughs.
Architecture explainers, stack walkthroughs, and shorter briefings on what changes when AI moves into enterprise operations.
Featured briefing
Inside the AI-Ready Enterprise Stack
A layer-by-layer walkthrough of the stack that has to exist before enterprise AI can do real work.
Recent briefings
Talks that extend the written work, not duplicate it.
About
Built across enterprise systems, data platforms, and applied AI.
Technical enough for architects and engineers. Clear enough for executive teams making operating bets.
I work where implementation detail meets business consequence: the infrastructure, runtimes, and operating patterns that make enterprise AI usable outside the demo.
The throughline is consistent. Identity, governance, integration, data platforms, runtime design, and execution all have to hold before AI becomes operationally real.