Data engineer, London.
The eight weeks nobody puts in the press release.
What the role is for, and who it is for.
You will build the data foundations under every Digital & AI engagement (ownership, definitions, quality gates, lineage) in the client's environment and repository, alongside their engineers. The role is measured on whether the model that follows holds a year later.
You will
- Rebuild field definitions, ownership and lineage from source to decision.
- Wire quality gates into the client's pipeline so bad data fails loudly.
- Pair with the client's engineers so the handover is a formality.
- Write the evaluation harness and agree thresholds before launch.
You have
- Shipped data pipelines that other people depended on, in a regulated setting.
- Opinions about definitions of 'customer' and the scars to back them.
- Comfort in someone else's codebase, under someone else's access controls.
If two of these are true and the third is close, write anyway.
What it is not.
A candidate who discovers these in week three has been sold a brochure. They are here so nobody is.
- A research role. You will not train novel models; you will make existing ones trustworthy.
- A greenfield role. Every environment is someone else's.
How we hire, in order.
Your CV and one page
The page is about a number you moved and how you know. We read every one; a partner, not a screener.
A 40-minute call
With the partner for the service. You ask most of the questions.
The case, on real data
Three hours with an anonymised ledger from a closed engagement. We want to see how you check a number, not whether you get the answer we got.
A day with the team
Paid. You sit in on live work and we sit in on you.
An offer in writing, within a week
From a published band. No negotiation theatre: the band is the band.