Datatonic Launches Updated “Turbo Templates” to Supercharge MLOps Adoption in 2024
London, 21st November 2023 – Datatonic announced today that it has released a new version of its open-source MLOps Turbo Templates, developed in partnership with Google Cloud’s Vertex Pipelines product team, to help data teams kickstart their Vertex AI MLOps initiatives. Customers like Delivery Hero, Vodafone, BT, and many more have already adopted MLOps and so can you.
Join today’s Google x Datatonic webinar to experience the MLOps Turbo Templates in action and continue reading to get started with MLOps.
What is MLOps?
MLOps encompasses the set of practices, technologies, and culture that enable Machine Learning to deliver real business value in an organisation. Borrowing concepts from DevOps and applying them to Machine Learning, MLOps enables teams to move rapidly from offline experimentation to a deployed production system and beyond, unlocking business potential from Data Science.
Even in the era of Generative AI, MLOps is a valuable investment for companies. Core MLOps capabilities for traditional Machine Learning can be used for Generative AI use cases. Watch the Vertex AI product team talking at Google Cloud Next 2023 about Building your MLOps strategy for Generative AI to understand common MLOps components and unique requirements for Generative AI.
What is Vertex AI?
Vertex AI is Google Cloud’s unified platform for AI + ML. It helps teams build, deploy, and scale ML systems faster, with tooling to support the entire ML lifecycle, from experimentation to production. Since its release in 2021, Vertex AI has continued to evolve, introducing powerful new features such as model monitoring, model versioning, and experiment tracking. Most recently, a new set of tooling has been released to support generative AI including Generative AI Studio, Model Garden, and Gen App Builder.
What are the “MLOps Turbo Templates” and why were they created?
The MLOps Turbo Templates are an open-source codebase that provides a reference implementation of the end-to-end ML lifecycle. Co-developed by Datatonic and Google Cloud’s Vertex Pipelines product team to help businesses kickstart their MLOps initiatives, they enable both Data Scientists and ML Engineers to use Google Cloud better and faster. Today, Datatonic is announcing a new version, bringing new features and capabilities to help organisations adopt MLOps practices on Google Cloud.
With more clients deploying ML models to production than ever before, our teams realised the power of MLOps solutions for maximising the value of clients’ ML models early on. Our MLOps solution for Vodafone, for example, has helped them significantly reduce time to production, and Delivery Hero’s MLOps Platform allows their teams to seamlessly leverage Machine Learning best practices.
The MLOps Turbo Templates combine Datatonic’s experience of implementing MLOps solutions with Google Cloud’s technical excellence to help reduce technical debt and adoption challenges many businesses face.
The templates can be used to create a production-ready MLOps solution on Google Cloud, covering the key components of a production ML system, including:
- Infrastructure-as-Code
- ML code versioning and deployment with CI/CD
- Data preprocessing
- Training a model
- Scheduled ML batch predictions
- Skew/drift detection
Although the Turbo Templates can be used in isolation for individual ML use cases, they can also form one of the key ingredients needed for adopting MLOps at scale across an organisation. The diagram below shows how a strong MLOps blueprint can fit into the wider picture of enterprise MLOps adoption.
What’s new in the MLOps Turbo Templates?
Since the initial release last year, the MLOps Turbo Templates have continued to evolve, building on Datatonic’s experience implementing MLOps with leading enterprises, as well as Google’s continued investment and innovation in state-of-the-art ML tooling. The new enhancements to the template cover two key themes:
- Adopting the latest and greatest in Google Cloud
- Improving the developer experience, making the template more lightweight and easy to use.
In terms of new functionality, the MLOps Turbo Template now uses Vertex AI’s model versioning to make managing your trained models much easier and neater, with model evaluations now viewable natively in the Vertex Model Registry.
Vertex AI Model Monitoring has replaced Tensorflow Data Validation to provide a built-in solution for monitoring data skew/drift, with added support for alerting. The Kubeflow Pipelines (KFP) SDK has been updated to the new KFP V2, which adds support for custom container components, pipelines-as-components, and the ability to store and version pipelines in Artifact Registry.
A lot of work has been done to improve the developer experience. By simplifying the pipelines and using more native Vertex AI functionality, the run time of the pipelines has been reduced by over 50%, allowing for faster iteration and less time spent waiting for pipelines to finish. The codebase has been significantly refactored, reducing the number of custom pipeline components and simplifying the approach to testing, both of which contribute to a lower barrier to entry and a reduced maintenance burden for teams using the template.
Lastly, the Infrastructure-as-Code has been rewritten from the ground up to provide full cloud infrastructure for dev, test and production environments using Terraform, making it even easier for Data Science teams to get started on Vertex AI.
Try it out for yourself!
The new release of the MLOps Turbo Template is now available in Datatonic’s open-source repository here (and Google Cloud’s repository in future), including all the documentation needed to help teams start with their use cases.
Get started today with the new three-part notebook series which will guide you through deploying your own MLOps project on Google Cloud and the productionisation of an ML model on Vertex AI in minutes.
Datatonic enables leading companies, like Vodafone, Sky, and Inmarsat, to make better business decisions with a powerful Modern Data Stack and accelerate impactful solutions with MLOps. For enquiries about new projects, get in touch.
