Machine learning
Machine learning for a defined task, tested honestly.
Machine learning models trained on your own operational data for a defined task, with honest testing and plain reporting. Part of our AI development work.
What we do
- We prepare your data, build and train machine learning models for a specific, agreed task, and measure how well they work on data they haven’t seen.
- We report results in plain language, including where a model is weak, and we tell you if a simpler rule-based approach would do the job.
- We package the model so it can run inside your own software or systems, with instructions for retraining it.
Typical projects
General examples of the kind of work we take on. They are not descriptions of past clients.
- Sorting documents, images or records into categories you define.
- Spotting unusual readings or events in operational data, such as equipment or system logs, for a person to review.
- Estimating future workload or stock levels from your own history, to help with planning.
- Suggesting related items within your own catalogue or content.
- Cleaning and preparing data so it can be used for reporting or later modelling.
Tech we use
- Python with pandas, scikit-learn and PyTorch.
- Notebooks and scripts kept in version control, so every result can be reproduced.
- Clear evaluation on held-out data, with the measures agreed at the start.
Your data
- We never use your data to train AI or machine learning models without your written consent, and only for your own project.
- Models trained on your data belong to you, and we don’t reuse your data or your models for anyone else.
More on our Privacy Notice.
What we don’t do
- We don’t build models that predict financial markets or prices, give investment or trading signals, or score people for credit or insurance.
- We don’t build models that make medical, legal or hiring decisions.
- We don’t guarantee a level of accuracy before we have seen and tested your data.
- We don’t give financial, investment, legal, tax or medical advice, and we don’t provide any regulated financial service.
- We never hold, receive or move money on behalf of clients or their customers.
Every project must also follow our Acceptable Use Policy.
How we work
Every project follows the same steps: a scoping call, a written fixed-scope quote with the price, milestones you review, and a handover with the source code, documentation and access. Ownership of the source code passes to you once the agreed price is paid in full.
Support is available Monday to Friday, 09:00–17:30 (our time zone, UTC+4), except public holidays, with no out-of-hours or on-call service.
See how every project works, what you get and how we handle your data
Who does the work
Our services are delivered by qualified independent contractors, under the supervision of our board.
All work is delivered remotely, on a professional services basis.
We work with small and medium-sized businesses and individual professionals buying for their professional work. We serve customers worldwide, except where we’re incorporated.
How to start
Tell us about your project with our enquiry form and say you’re interested in machine learning. Already a customer? Open a ticket instead.