Company · Careers

Engineers who shipsystems that run unattended.

The work here is building AI that finishes real work for real companies, and then being the one who answers when it does not. If you would rather ship something that runs than demonstrate something that might, read on.

A robot arm on a desk in an office

The work

Small teams, whole workflows.

An engineer here scopes, builds and hands over the same work. You sit with the people who do a job today, map it, build the models and agents that do it, prove it against their real records, and hand it to them with the runbooks. Then you are on call for what you built.

There is no separate sales layer, no research team that never meets a customer, and no operations team that inherits what someone else shipped. The engagement is yours from the map to the hand-over, and the company's own operations run on the same systems you build.

The work is technical and it is close to people. You will read a warehouse's despatch notes and a clinic's intake forms, and you will explain to the person who fills them in why the system did what it did.

What it asks

What the work needs from you.

  1. 01

    You finish things

    A workflow that is mostly automated still needs a person to watch it. The job is the last twenty percent: the exceptions, the edge cases, the hand-over.

  2. 02

    You measure before you believe

    You agree on how to tell whether the system is right before you build it, and you are the first to say when it is not.

  3. 03

    You can sit in the room

    Mapping a workflow means listening to the people who do it, in their words, and being straight with them about what will and will not change.

  4. 04

    You work across the stack

    Models, agents, integrations, the ledger underneath, and the deployment in a customer's own account. Nobody owns one layer.

  5. 05

    You write it down

    The map, the evaluations, the runbooks and the reasons for every decision are deliverables, not afterthoughts.

  6. 06

    You answer the call

    What you built runs unattended until it does not. When it does not, you are the person who knows why.

Where we need people

The kinds of work.

There is no list of open roles. Tell us which of these you do and what you have shipped, and we will tell you whether there is a place for it now.

  • 01

    Models

    Training, tuning and evaluating models on a company's own data, and serving them inside its account.

  • 02

    Agents and automation

    Agents that run long tasks across several systems with one person's permissions, and the integrations they act through.

  • 03

    Documents, speech and vision

    Reading forms and scans into records, voice that books and confirms, cameras that count and inspect.

  • 04

    Data and MLOps

    The pipelines, evaluation harnesses and monitoring that keep a shipped system right as the data drifts.

  • 05

    SoloOne

    The suite itself: eight modules on one ledger, and the agents that work inside them.

  • 06

    Forward-deployed engineering

    Engineers who spend the engagement with the customer, mapping the work and shipping it in place.

How hiring runs

Four steps, no puzzles.

We hire the way we work: on real work, measured, with the people you would be working with.

  1. 01

    Write to us

    Use the contact page. Say which kind of work you do and point us at something you shipped that runs on its own.

  2. 02

    A conversation

    An engineer talks with you about that work: what it does, what went wrong, how you knew, and what you would do differently.

  3. 03

    A piece of real work

    A small, paid piece of work of the kind we do, on our own systems, with an engineer alongside. You see the job; we see how you do it.

  4. 04

    An answer

    A plain yes or no from the people you would work with, with the reasons, whichever way it goes.

Questions

Asked before applying.

Where are the open roles?

There is no list. The areas above are the kinds of work we do; if you do one of them, write to us with something you have shipped and we will say plainly whether there is a place for it now.

Do I need a background in machine learning?

For the model work, yes. For the rest, what matters is that you have shipped systems that run without you, that you measure before you believe, and that you can sit in a room with the people whose work you are changing.

What does the paid piece of work involve?

A small task of the kind we do every week, on our own systems, with an engineer alongside. It is paid, it is scoped to be finished, and it is the same work you would be doing if you joined.

How do I apply?

Through the contact page. Say which area you work in, link to something you have shipped, and an engineer will reply.

Start

Tell us what your teamstill does by hand.

We map the workflow with you and say plainly what can run on its own, what should not, and what it takes.