Scaling agentic architectures in regulated industries
Keynote at Digital Day, Belgrade. How regulated industries can ship enterprise AI agents without giving up compliance, control, or security.
Writing · Talks
Separating the language from the execution.
Essays and talks on AI agents — and the unglamorous engineering that keeps them working in production.
Doing chatbots since before they were good.
Selected writing & talks Full archive →
Multi-agents are secretly distributed monoliths
Why most "multi-agent" systems are orchestration dressed up as autonomy — and the real engineering problem hiding behind the hype.
Orchestrating A2A and MCP with Rasa
Wiring agent-to-agent and MCP into production agent systems — without handing runtime control to an LLM and hoping.
From NLP to LLMs: taking AI agents to production
On the a16z Podcast with Martin Casado — a decade of lessons on what it actually takes to run AI agents in production at enterprises.
I'm so glad we got rid of intents.