Senior engineer, avoiding AI roles because you never did ML?
AI engineers don’t build the brain. Here’s what they actually do.
On a call this week I spoke with a senior architect — 20+ years across Java, cloud, and microservices — who is actively moving into AI engineering.
Even he had the map of these roles wrong.
He assumed part of the field was off-limits: the “low-level” side, where engineers build the models themselves. Territory for a different breed of engineer.
Here’s the picture I drew for him.
Training large language models — building the brain — is research work. It happens inside a handful of frontier labs, done by PhD teams. Almost no company on earth hires for it.
Every AI role you’ll actually see posted is about using the brain. The model runs on the frontier labs’ servers. You work with it through API calls.
Two roles live in that territory, and both are hiring:
1. AI engineer.
Works inside a regular company. Builds AI into the business — RAG over company documents, agents that automate workflows, evals that keep it all honest. Every piece built on API calls to a model someone else trained.
2. Forward-Deployed Engineer.
Works for an AI product company. Sits with a client and adapts their employer’s AI product to that client’s business. Half technical, half client-facing.
Same core skill set in both seats. Neither role trains a model. Both need someone who can build production software around a model that already exists.
That is why a 20-year architect walks into this field with an advantage over a fresh ML graduate. The work is engineering — systems, integration, production. The AI layer stacks on top of what he’s done for two decades.
So before you filter yourself out of another AI engineering listing because your resume says “systems” instead of “machine learning” — read the listing again. It’s asking for a builder.
Kirill
P.S. The brain is already built. We teach everything around it. See if you qualify for the AI Engineering Program!