Senior AI Engineer · Copenhagen
Building the agent is one part of the job. Getting a whole company to use AI well is the rest.
I'm Bhanu Chaddha. I lead the building of AI agents and an AI platform inside a large global enterprise, and I drive AI adoption across the organization. I work as a forward deployed engineer, so I own the full path: finding the right problem, building the agent, evaluating it, and running it in production.
I also speak and write about adopting AI well: using its full potential, and using it safely.
- 60
- posts published
- 11 / 21
- Agents in Production
- 7 / 12
- Structured Thinking
- 2 / 12
- The Delivery Layer
- 1
- The AI Agent Playbook
More than 13 years of building software, the last few on AI agents. Invite me to speak
What I do
AI agents in production
The full path inside a large enterprise: finding the right problem, building the agent, evaluating it, and running it.
Agents in Production series →
Coding agents at work
How I work with AI coding agents every day: the setup, the guardrails, and the rules that stop the same mistake happening twice.
Free setups and checklists →
AI adoption and teams
Leading an AI developer community of 750 people, training leaders and their teams, and working out what changes when a whole team builds with agents.
The Delivery Layer series →
Speaking and writing
Talks on what happens to teams when code gets cheap, a book on building agents, and writing that says what is real.
Talks and speaker bio →
Recent posts
View all →Claude Code Guardrails: Permissions and the Sandbox
Claude Code runs with your keys. What it can reach by default, which settings.json rules help, why they are not a wall, and how to turn on the built-in sandbox.
Say It Back Before You Leave the Room
Structured Thinking · part 7
Notes preserve a conversation but never structure it. Compress a rambling brief into three named themes and say them out loud while it can still be corrected.
Precise Enough to Be Wrong
Structured Thinking · part 6
Precision is a resource, not a virtue. Decide how exact the answer needs to be before you start, or the deadline decides it for you at the worst moment.
Guess First, Then Try to Prove Yourself Wrong
Structured Thinking · part 5
Analysis with no hypothesis expands to fill the time and yields a folder, not an answer. Commit to a guess on day one, then spend the time killing it.
What I will argue for
Browse by topicAI Agents 15RAG & LLMs 19Docker & Kubernetes 14Engineering practice 15All topics →
Looking for a speaker, or want to compare notes on any of this? I'm happy to talk.