Now booking Q1 2027
Now booking Q1 2027Start a project

15:00 — this craft's hour on the studio wall

Machine learning development for startups and product teams

We take machine learning from a question about your data to a model running in your product. Every project starts by measuring whether a model can beat the simple approach you use today, and only goes further if it can.

Start a machine-learning projectEstimate the cost

What we help with

  • You have data, but no idea if it's enough

    We audit what you have, how clean it is and what it can realistically predict, and give you a straight answer before you pay for a full build.

  • A model that works in a notebook but not in production

    We turn it into a versioned, tested service with monitoring, so it runs on fresh data and someone notices when it drifts.

  • Forecasts and recommendations made by gut feel

    Demand, churn, pricing, the next product to suggest: we model the decisions you already make by hand, and compare the model against how they are made today.

  • Text your team reads all day

    Classification, extraction and search over support tickets, documents or reviews, using language models where they help and simpler models where those are enough.

What you get

  • A data audit and a written assessment of what is feasible
  • A baseline to beat, so any improvement is measured, not assumed
  • Trained models with an evaluation report in plain language
  • An API or batch pipeline that serves predictions to your product
  • Monitoring for data drift and model quality after launch
  • Model weights, training code, datasets and experiment history, in your accounts

What we build with

Mainstream, open tools, so the work doesn't depend on us.

Python, pandas and scikit-learn
often the right answer; gradient-boosted trees are hard to beat on tabular business data.
PyTorch
for language, sequence and image models, when the problem needs them.
MLflow
experiment tracking and a model registry, so every result can be reproduced.
Feature pipelines
scheduled jobs that keep training and serving data consistent, on the cloud you already use.

How an engagement runs

Most projects move through these in order. You can stop after any of them and keep what was made.

  1. Scope sprint

    2 weeks · fixed fee

    Turn an idea into a plan you can trust before you commit to a build.

    A data audit, a first baseline and a plan with a fixed quote.

    • A clickable prototype
    • A list of risks and how to retire them
    • A fixed quote for the build
    • The fee is credited against the build

    Sample price: from $3,000

    Start with a scope sprint

  2. Build

    6–12 weeks

    We design and ship the product, with something real to try every Friday.

    Model development and production integration, with an evaluation you can read every week.

    • Weekly builds you can open on your phone
    • Design, engineering and launch
    • Code, models and data handed over
    • 30 days of fixes after launch

    Sample price: typically $15,000–60,000

    Plan a build

  3. Ongoing team or ML pilot

    monthly

    A small senior team that keeps improving a product you already run.

    A pilot first when the question is open, to measure whether a model is good enough; then retraining and monitoring.

    • A fixed team, no rotating juniors
    • A monthly plan and a weekly demo
    • Pause or stop with a month's notice

    Sample price: from $4,000 a month

    Talk about a team

Prices here are samples, to give a sense of scale. A scope sprint ends with a fixed quote for your project. For a rough range before you write, try the cost estimator.

Common questions

How much data do we need?

It depends on the problem and how noisy the data is. Rather than quote a number, we look at what you have, train a quick baseline and tell you whether more data would help, and roughly how much.

What if the model isn't accurate enough?

Then you find out early, in the pilot, before anything is built around it. We agree with you at the start what good enough means, and if the model does not reach it we say so and recommend what to do instead.

Can you work with our existing data team?

Yes. We can pair with your engineers, work inside your repositories and cloud, and hand over in a way your team can carry on from.

Where does our data go?

It stays in your cloud accounts wherever possible. We sign an NDA before you share anything, and copy data out only with your agreement.