Start a conversation

Building machine learning and AI models

Useful AI starts with a well-defined problem and honest data. We build machine learning and AI models for specific business requirements, following a disciplined process from data gathering to prediction — and we're explicit about what a model can and can't be trusted to do.

Discuss your project See how it works

What we do

Built for one requirement

We start from the decision the model will support and the data available to support it, not from a favourite algorithm.

Honest about uncertainty

Every model makes mistakes. We report how a model performs, where it is weak and what it should not be used for, so decisions built on it are informed ones.

Techniques

Deep learning and neural networks, regression, decision trees and clustering — chosen to fit the problem, the data and how results will be used.

How a model gets built

The same seven stages, whatever the algorithm.

  1. Data gathering

    Identify and collect the data that describes the problem.

  2. Data preparation

    Clean, combine and shape it — often the largest share of the work.

  3. Choosing a model

    Select approaches suited to the problem and the way results will be used.

  4. Training

    Fit candidate models on the prepared data.

  5. Evaluation

    Measure performance on data the model hasn't seen, and study where it fails.

  6. Parameter tuning

    Adjust and re-test until results are good enough for the intended use.

  7. Prediction

    Put the model to work inside an application, a report or an API.

Typical applications

Classification

  • Fraud detection
  • Image classification
  • Customer retention

Regression

  • Market forecasting
  • Growth predictions

Clustering

  • Customer segmentation
  • Target marketing
  • Recommendation systems

Talk to us about AI and machine learning models.

Tell us what you're working with and where it hurts. We'll tell you honestly whether and how we can help.

Discuss your project

Or email [email protected]