Senior AI Engineer

About the role

Most of what we ship with AI in it is a retrieval problem, an evaluation problem and a permissions problem with a model on top. You will work on client systems, often in regulated sectors, where somebody has to be able to explain months later what the system saw and what it decided. If your instinct on hearing that is to build an evaluation harness before writing a prompt, this is the right role.

What you will do

• Design and build retrieval pipelines over client data, including the unglamorous work of getting the index right before touching the model.
• Write evaluation harnesses that turn a client specification into a score, so that done means something you can point at.
• Own an AI feature end to end: data boundaries, inference cost, latency budget, failure behaviour and rollback.
• Join client calls and explain in plain language what the system will and will not do. Sometimes you will be the person saying no.
• Review other engineers’ work and have yours reviewed. Nothing merges on one pair of eyes.

What we need from you

• You have put something with a model in it into production and lived with it afterwards.
• Strong Python, and enough backend engineering to own the service the model sits behind.
• You can explain why a retrieval change beat a model change on a system you actually worked on.
• You are comfortable with the compliance conversation: what data may leave the environment, and what has to be logged.
• You write clearly. Much of this job is making a technical position clear to a non-technical client.

Useful but not required

• Experience with regulated data, particularly HIPAA or PCI scope.
• You have run a model self-hosted and know what that actually costs.
• Open source work, a paper, or a write-up we can read.