MLOps

MLOPs

There are a number of barriers that prevent companies from successfully implementing data and ML projects. It’s generally considered to be significantly harder than implementing software projects due to the cross functional complexity of data and ML pipelines. As a result, even if you hire the best data scientists, their work often fails and many companies give-up on their projects before they’ve begun to see the value of the technologies.

Solving these challenges is a new field of expertise called MLOps

We are working to incorporate MLOps into Quix so that your data team has a seamless journey from concept to production. The key steps are:

Discover and access data

Any member of any team can quickly access data in the Catalogue without support from software or regulatory teams.

Develop features in historic data

Use Visualise to discover, segment, label and store significant features in the catalogue.

Build & train models on historic data

Use Develop and Deploy to:

• Write model code in Python using their favourite IDE.

• Train models on historic data.

• Evaluate results against raw data and results from other models.

• Rapidly iterate models with GIT version control.

Test models on live data

Connect models to live input topics to test them against live data sources. Review the results in Visualise.

Build a production pipeline

Use Develop and Deploy to:

• Connect validated models to live output topics.

• Daisy-chain models using input and output topics.

• Work seamlessly with engineers to hook-up software services.

Deploy production models

With one click, data engineers can deploy their Python models to production without support from software engineering or DevOps teams.

Monitor production models

Data teams can:

• Ensure that components in a production pipeline operate correctly through the product lifecycle.

• Build and deploy services that detect data drift or unexpected results.