Client delivery
NoCodeML — automated machine learning from data to API.
Zero One Touch developed NoCodeML for Univera as a custom automated machine-learning platform. Users supply data; the platform analyses it, evaluates model approaches and trains candidate models, then exposes the selected model through an API.
The work behind a model API
Making a model available through an API involves more than a training run. The data needs to be analysed, candidate approaches evaluated and model parameters optimised before a version is selected for use.
We built NoCodeML to carry out that sequence within a single platform. The user provides the data rather than assembling and configuring each stage of the machine-learning pipeline.
How the platform handles supplied data
- Analyse the supplied data.
- Evaluate model approaches and optimise ML parameters.
- Train candidate models and prepare them for production use.
- Expose the selected model through an API.
When the data pool changes
An update to the underlying data pool can trigger another training cycle. NoCodeML compares the new candidate’s performance with the model currently published and can publish the improved version when the configured success criteria are satisfied.
The comparison and the publication criteria matter: a new training run does not automatically replace the model already in use.
What this delivery demonstrates
NoCodeML connects automated model training, retraining and performance comparison with API serving. It is an example of our ML and platform-engineering work for a client, rather than a public Zero One Touch subscription product.
WORK WITH US
Discuss a related project
Tell us about the data, application or operating environment you need to work with.