I build AI systems that have to survive contact with production

I’m Thomas De Vos, a principal AI architect and cloud engineering lead with more than 25 years in software delivery, financial services, cloud platforms, and AI engineering.

These days my work is mostly about agentic AI: taking the exciting part of a demo and turning it into something a real team can run, review, observe, secure, and improve. That usually means less hand-waving and more engineering: permissions, evals, tool boundaries, CI/CD, observability, cost control, and human review.

That is also what I write about here.

What I write about

This site is for engineers, architects, technical leads, and product-minded builders who are trying to make sense of AI beyond the hype cycle.

I write about:

  • Claude Code and agentic coding workflows
  • production AI agents and LLM systems
  • software architecture for AI-enabled teams
  • cloud engineering, automation, and delivery
  • the messy gap between prototype and production

I am interested in what works when there are real users, real security constraints, real budgets, and real people on call.

My book on Claude Code

Cover of Claude Code: Building Production Agents That Actually Scale by Thomas De Vos

My book Claude Code: Building Production Agents That Actually Scale is now live on Amazon Kindle.

The book is a practical guide to using Claude Code in production engineering environments. It covers agent loops, tools, hooks, MCP, permissions, evals, observability, cost engineering, and the human review practices that keep agentic workflows from becoming a black box.

If you are experimenting with Claude Code and wondering how to turn it into something your team can trust, that is the problem the book is trying to solve.

Read the book on Amazon Kindle

A bit more personal

I have completed four Ironman triathlons. That part of my life still shapes how I think about engineering: steady practice, honest feedback, recovery when things go wrong, and patience when the results take longer than expected.

I also have a soft spot for Prince, Kung Fu discipline, and travel with my wife Marilin. Those may not sound related to AI engineering, but they are part of the same thread for me: craft, rhythm, repetition, and staying curious.

Contact

You can email me at thomas.devos@gmail.com, find me on LinkedIn, or follow my writing on X.