Canonical’s AI and ML solutions feature…

Architectural freedom. Fully automated operations. Accelerated Deep Learning

Canonical’s AI solutions such as Kubeflow on Ubuntu use your existing on-premise clusters and GPUs efficiently, giving you architectural freedom with storage and networking while sharing operational code with a large community. From your developer workstation, to your racks, to the public cloud, AI on Ubuntu is accelerated with the latest tools, drivers and libraries.

The standard for enterprise machine learning, from Silicon Valley to Wall Street, for the Fortune 50 and for startups.

Contact us for machine learning, deep learning and AI consulting ›

Private cloud and HPC architecture

GPU acceleration of AI and machine learning workloads requires careful configuration of the underlying hardware and host OS. Canonical’s Ubuntu is the leading platform for public cloud GPU instances and Canonical offers private cloud expertise to match.

Build a GPU cluster and share it with multiple tenants using Canonical OpenStack - then operate Kubernetes on top for HPC and high-throughput AI / ML data science.

Start learning about AI with Kaggle

Kaggle competitions are a great way to start learning about AI and develop your skills. For beginners, consider starting with one of the following previous competitions:

Effective decision making

With deep learning on vast amounts of data, make quicker and more effective decisions. Over time, the machine algorithms learn to distinguish what data is important what isn’t. Insight extracted from AI will allow you to optimize your processes.

Operational predictions improve SLA

Using real-time telemetry data from the infrastructure in your data center, from hardware to software assets, leverage AI to help predict when components will fail or need to be regenerated. This can help you uphold impressive service availability metrics.


Kubeflow helps you build composable, portable, and scalable machine learning stacks. With Kubeflow you can speed up the AI tools and framework installation process, particularly leveraging GPUs from Nvidia.

Without Kubeflow, building production-ready machine learning stacks can involve a lot of infrastructure and devops work — mixing components and solutions, wiring them together, and managing them. This complexity can be a barrier to adopting machine learning, and it can significantly delay achieving the business benefits you are hoping to receive. And then you want to launch something production worthy; start all over again.

Kubeflow solves these challenges by pulling together a handful of technologies and components that let you get a stack up and running quickly. You can accelerate that roadmap and benefit from community and/or corporate support.


TensorFlow is an open source software library for high-performance numerical computation. Its flexible architecture allows easy deployment of computation across a variety of platforms (CPUs, GPUs, TPUs), and from desktops to clusters of 乐天堂fun88备用网址s to mobile and edge devices. Originally developed by researchers and engineers from the Google Brain team within Google’s AI organization, it comes with strong support for machine learning and deep learning and the flexible numerical computation core is used across many other scientific domains.

TensorFlow comes with visualisation technology — TensorBoard. It features graphs, histograms and helps with visualising learning.

Tensorflow from Google is officially published for Ubuntu


With you can create a multi-user Hub which spawns, manages, and proxies multiple instances of the single-user 乐天堂fun88备用网址.

Project Jupyter created JupyterHub to support many users. The Hub can offer notebook 乐天堂fun88备用网址s to a class of students, a corporate data science workgroup, a scientific research project, or a high performance computing group.

Operations automation

The real challenge of Kubeflow is everyday operations automation, year after year, while Kubeflow continues to evolve rapidly. This includes the everyday automation of the stack under Kubeflow. Canonical solves this problem with model-driven operations that decouple your architectural choices from the operations codebase that supports upgrades, scaling, integration and security.

Total automation of GPU enabled infrastructure

Eliminate the extra steps needed to take advantage of your GPU’s by leveraging Kubeflow. With drivers tailored to your chipset, you’ll get the most out of your investment, and speed up your deep learning initiatives.

Artificial Intelligence infrastructure architecture

To get the most out of Kubeflow, you’ll want to run it on an effective supporting stack. Minimally, leveraging Charmed Kubernetes gives you the benefits of perfect portability between your private data-center and the public cloud. Charmed Kubernetes on Canonical Openstack unlocks further benefits, as described below.


Every ounce of performance matters. If you’re building a private cloud you want the maximum performance for your workloads, the maximum utilisation in your data center, and the maximum economic efficiency. Canonical delivers all three.


Storage performance and economics are tricky to balance in a cloud environment. Canonical will help you architect your storage across the cluster to balance price and performance, ensuring the right mix of resilience, latency, iops and integrity for your particular deployment.


Network performance is critical for speeding up large deep learning exercises. The major factor in perceived cloud performance is aggregate network throughput and latency across the underlying cluster. Canonical’s work with hyperscale public clouds ensures that we have deep insight into the dynamics of cloud network performance and security best practices for large-scale multi-tenanted operations. Our work with telco groups for NFV and edge clouds ensures that we can work well in complex environments where latency and security are critical.

Operational Dashboards

Operations in highly coherent large-scale distributed clusters require a new level of operational monitoring and observability. Canonical delivers a standardised set of open source log aggregation and systems monitoring dashboards with every cloud, using Prometheus, the Elastic Search and Kibana stack (ELK), and Nagios.

Operational dashboards

These dashboards can be customised or integrated into existing monitoring systems at your business.

Learn more about AI/ML and Kubeflow


A detailed look into the AI and ML landscape, how to deploy your first model and more.


Articles from across the web on getting started with AI and Kubeflow in your workplace.


Examine the fundamentals of a successful AI project that helps your organisation achieve their AI ambitions.

Get started with enterprise AI ›

Get the most from your workloads

Find out why Ubuntu is the standard for enterprise machine learning for Fortune 50 companies and for startups.

Get in touch