The interesting problem in applied AI right now is not capability. It is constraint. What is authorized in the environment you are deploying into, what data can legally reach the model, what happens when it is wrong, and who is accountable when it is.
We work with organizations that have those constraints in a sharper form than most — regulated environments, public sector data, and small teams with no capacity to operate something fragile. The work is building things that hold up in those conditions.
What you would work on
Designing and building retrieval and agentic systems that work against a client’s real data rather than a demo set.
Working inside authorization boundaries. Knowing which models are available in which cloud at which impact level is part of the job, not a footnote.
Evaluation: establishing whether a system actually performs before it goes anywhere near a user.
Guardrails, monitoring, and the operational layer that makes a model deployment supportable by people who did not build it.
Explaining all of the above to a client who is being sold something simpler by someone else.
What we look for
You have put something using a model into production and dealt with the consequences.
You are sceptical of your own results and you test them.
Python, and familiarity with at least one major cloud’s AI services — Bedrock, Azure OpenAI, Vertex.
An interest in the governance side rather than an aversion to it. On our engagements it is usually the hard part.
Other roles
BUSINESS ANALYSIS
Business Analyst
Turning what a client says they need into something that can be built and bought.
INFRASTRUCTURE
IT Specialist
Hands on the systems. Identity, endpoints, networks, and the day-to-day that has to work.
DATA
Data Scientist
Finding what the data actually supports, and saying so plainly.
Register your interest
No cover letter. We will read it and we will reply either way.