Work with us
Tell us what you need to build, evaluate or decide
Enquiry aboutAI agent engineeringPlanned
This service line is planned, not yet offered — an enquiry registers interest. The form below is pre-set to it.
Audiences
Who this is for
Teams building or buying AI
Startups, product teams and enterprises that need an AI feature evaluated, built or made more reliable — or a vendor choice backed by evidence.
AI vendors & labs
Model and tool builders who want task-specific evaluation, adversarial testing or an independent reading of benchmark results, with published settings.
Researchers & educators
Academic and independent researchers interested in reproducible evaluation, benchmark analysis or clear technical explainers.
Here to learn? Start with paper breakdowns and insights, or get new work by email through the newsletter.
Process
How it works
Enquiry
You describe the problem, the decision you need to make and any constraints — through the form on Work with us.
Scoping conversation
We talk through the system, the data and what success looks like, and decide together whether we are the right fit.
Written proposal
Scope, deliverables and acceptance criteria are written down and agreed before any work starts.
Delivery with evaluation built in
Work is measured against the agreed acceptance criteria as it progresses, with results you can inspect rather than demos.
Handover
Deliverables, documentation and the means to re-run the evaluations pass to your team, so the work can be checked without us.
This is the process we intend to follow. We do not publish project timelines or delivery service levels until we can commit to them; anything time-bound is agreed in the written proposal.
Your brief
What to include in a brief
The decision or outcome
What the work must decide or deliver — choosing a model, shipping a feature, passing a release gate.
The system and the data
What exists today and what kind of data is involved. Describe it in general terms; don’t attach it.
How success will be judged
Metrics, acceptance criteria or the threshold that would change your decision, if you know them.
Constraints
Deployment environment, data residency, access control, and any compliance requirements.
Timing and budget
Fixed dates and a budget range if you have them. “Not decided yet” is a perfectly good answer.
Who is involved
Who makes the decision, and who would work with us day to day.
FAQ
Questions
Are your services available now?
Not yet. Every service line is currently marked Planned: we are designing them and are not taking engagements. An enquiry registers interest and creates no obligation on either side. Each service page shows its current status.
Do you publish prices or timelines?
No. Scope, deliverables, acceptance criteria and any timing are agreed in a written proposal before work starts. We do not publish delivery service levels we have not committed to.
Can you sign an NDA?
If your work needs one, say so in your enquiry. Until an NDA is signed, please share only information that is not confidential — no confidentiality obligation arises before then.
We build a model or AI product. Can you evaluate it independently?
You can ask: choose “Independent evaluation of our model or product” as the enquiry type. Anything we publish labels each number by its source (vendor-reported, third-party or HARBAN internal), as described in our benchmark methodology.
I’m learning about AI — is this form for me?
This form is for projects, collaborations, evaluations, partnerships and media. For learning, start with our paper breakdowns and insights, or subscribe to the newsletter.
What happens after I send an enquiry?
You will see a confirmation page with a reference for your enquiry. If it looks like a fit, the next step is a scoping conversation. If it is not something we can help with, we aim to say so plainly. We aim to reply within 24 hours.