About
The longer story.
I’m Caner Yilmaz, a Principal Platform Engineer specialised in Agentic AI. For over ten years I’ve built platforms and applications in environments where failure is not an option: from the Ministry of Defence and the Council for the Judiciary to a global scanning platform processing hundreds of thousands of datapoints per second.
The common thread: building things that don’t exist yet, and modernising complex technical environments. Designing and realising complete platforms from scratch — from database to frontend to deployment — and integrating security, auditability and resilience from the start, never afterwards.
In recent years my focus has been Agentic AI in real engineering workflows. At USNL, the company I co-founded, I built a multi-tenant AI platform where agents pick up tickets, write code, test their own output and submit merge requests. At Modat I converted an entire engineering team to parallel agentic workstreams. Not as an experiment — in production.
Alongside that, I work as a freelance engineer: for organisations that want a platform or application built from scratch, want to modernise their existing platform, or want to take AI seriously into production.
Engineering philosophy
- Simple wins. A component has a clear place and lifecycle in the architecture.
- The right context at the right moment matters more than as much context as possible.
- Agents must be able to test and verify their own output.
- When agents make structural mistakes, improve the knowledge base, skill or workflow — not the individual output.
- Risk doesn’t have to be zero, but it must be visible and consciously accepted or mitigated.
- Don’t migrate for the sake of migrating.