Careers

We hire people who have shipped it

A small senior team at our Dhaka development center, doing AI engineering for clients across the US, the UK, Europe and Australia. We hire slowly, we pay for depth, and we would rather leave a seat empty than fill it badly.

No posted roles right now, and we are still reading applications

We are a small team and we hire when the work demands it rather than on a schedule, so there is nothing posted today. We do keep a list, and the last two people we spoke to seriously came from open applications rather than adverts. If the way we work sounds right, send us something you have built.

hello@opendimension.ai, with something you made rather than a covering letter.

How we work

What it is actually like here

Written so you can tell whether you would hate it.

01

You will talk to clients

We do not put a layer of account management between engineers and the people whose problem they are solving. If you would rather never be on a call, this is the wrong team. If you would rather hear the problem directly than receive it as a ticket, it is the right one.

02

The work is unglamorous more often than not

A realistic split is that most of an engagement is content auditing, document extraction, permissions and evaluation, and a minority is the part that looks like AI in a demo. People who enjoy the plumbing do well here. People who want to spend their time on model selection do not.

03

Writing matters as much as code

Handover documents, evaluation reports, and the note that tells a client we do not think they should build the thing they asked for. If you cannot write clearly you will find this frustrating, because a lot of the value we deliver arrives as prose.

04

Overlap hours, not night shifts

Clients are in the US, the UK, the EU and Australia, so some overlap is real and we agree it per engagement. We are not running a night shift and we do not pretend the timezone difference is nothing.

05

Senior means judgement, not years

We care whether you can tell when a project should not go ahead, whether you notice that the evaluation set is leaking, whether you push back on a scope that will not work. That is what we are hiring for and what the interview is about.

The process

How hiring runs

Four steps, roughly two weeks, and we tell you where you stand at each one.

01

Send something you built

A repository, a write up, a system you are proud of. We would rather read your work than your CV, and a paragraph on what was hard about it beats a covering letter.

02

A conversation, not a quiz

Forty five minutes on something you have shipped: what broke, what you would do differently, where you were wrong. No algorithm puzzles.

03

A paid exercise

A small realistic problem, time boxed, and we pay for your time. We will not ask you to work for free and we will not ask for a week of it.

04

Meet the team and ask us hard things

You should interrogate us as much as we interrogate you, including about money, capacity and how stable the pipeline is. We will answer straight.

What we look for

The disciplines we hire into

You do not need all of these. Depth in one column and working knowledge of another is the usual shape.

AI engineering

  • Retrieval and ranking
  • Agent design and tool use
  • Evaluation harnesses
  • Guardrails and injection defence
  • Fine tuning and distillation

Data engineering

  • Ingestion and change capture
  • Document extraction
  • Knowledge graphs
  • Index design
  • Redaction and residency

Platform and reliability

  • Cloud architecture
  • Kubernetes and serverless
  • Observability and tracing
  • Incident response
  • Cost optimisation

Delivery

  • Workflow mapping
  • Solution design
  • Technical writing
  • Client engineering
  • Handover and training

Send us something you made

Open applications are read properly, including when nothing is posted. Tell us what you built and what was hard about it.