Did you miss CNCF’s Kubernetes AI Day at KubeCon North America 2021 back in October? If you did, you may have missed that all the talks from the event have been uploaded to YouTube. In part one of this two part blog series, we’ll give you an executive summary of the first five talks of the day.
Kubeflow Pipelines v2: The Next Generation of MLOps on Kubernetes
Karthik Ramachandran, Product Manager of Google Cloud Vertex AI – Google, gave a talk to introduce Kubeflow Pipelines v2. The goal with v2 is to make it significantly easier to author and manage model training Pipelines. He reviewed for us the new capabilities of the project, a little bit about the plans for the future, plus discussed where and how the community can pitch in.
- What are Kubeflow Pipelines?
- An overview of Kubeflow Pipelines v2 including metadata and metrics, streamlined DSL, UI, the “IR”
- Metadata and metrics will deliver a standardized type ontology, tools for querying and updating metadata, ability to query and visualize artifact lineage and finally a set of well defined integration points for app developers
- A simplified SDK providing more expressive tools for building components and pipelines
- An improved UI to support larger more complex workflows, a dynamic visualization of all control-flow features, ability to scale to “hundreds of steps” and finally provides improved visualizations of metadata and metrics.
- An open source, platform independent representation of pipelines that makes it easy for tools and developers to build new SDKs, editing tools and execution engines
Evolving with Kubernetes: Embracing Model Ops
Next up, Steven Huels, Sr Director of AI Cloud Services at Red Hat gave a talk on the advantages of applying DevOps principles to AI and Machine Learning to help you get your models from pilot to production safely, securely, and with repeatability.
- Data and AI workloads are gaining popularity on Kubernetes
- What are the challenges for AI/ML projects today
- Operationalizing AI/ML is not trivial
- Kubernetes and DevOps based platforms are helping address these challenges
- ModelOps with a Kubernetes and DevOps powered Hybrid Cloud Platform
What, Why and How of Federated Machine Learning – Implementation Using FATE, KubeFATE, and FATE-Operator for Kubeflow
- FML makes it possible to extract insights from widely dispersed data sources. It may also be used with privacy enhancing measures to reveal insights without revealing the underlying data.
- It gives global organizations an opportunity to utilize worldwide data while adhering to regulations for different geographic regions, allows insights to be developed without moving large amounts of data to a central location, and in a multi-party scenario allows each party to control their individual decision to participate.
- Introduction to FML discussing its basic principles, its manifestations, high-level use-cases, and tie these in with the technology being developed.
- A demonstration of how FML can be implemented on Kubernetes using FATE, KubeFATE, and integrated with Kubeflow using the FATE-Operator.
- New features being implemented into FATE for realistic use-cases.
Security Best Practices for AI on Kubernetes
In this talk, Guy Salton of Run:AI, presented the security concerns for AI workloads running on Kubernetes and how to mitigate them.
- Containers and Kubernetes for Data Scientists
- What data scientists need
- Security concerns for IT/Security on Kubernetes
- How to make everything more secure
Making Complex R Forecast Applications Into Production Using Argo Workflow
In this talk, Pedro Szloma Herr Zaterka, Matheus Sesso Gay and Natalia Costa Araujo of 4intelligence presented on the topic of how processes developed in R are created for problems restricted to academics researchers. However, as companies search for data and scientific methods to guide business decisions, creating a scalable R environment will be a critical step towards success. This talk explores that journey to cloud, microservices and beyond.
- Main challenges with R including bringing it to production, optimizing cost efficiency, allowing for faster and simpler deployments and the reliability of the results
- R initial pipelines and code sharing in GitHub
- Migrating R to the cloud including “Lift and Shift”, using R in the cloud and monolithic structures
- Introducing the idea of “Forecast as a Service”
- The pros and cons of R
- An overview of the architecture of how to bring R to the cloud including decoupling and scalability challenges
- Dockerization of R
- A recap of trial and errors, lesson learned
- Leveraging Argo Workflows with R
- Enabling a microservices-based architecture
- Achieving massive scale
Stay tuned for part 2 of this blog series where we recap the rest of the talks presented at Kubernetes AI Day North America 2021.
Book a FREE Kubeflow and MLOps workshop
This FREE virtual workshop is designed with data scientists, machine learning developers, DevOps engineers and infrastructure operators in mind. The workshop covers basic and advanced topics related to Kubeflow, MiniKF, Rok, Katib and KFServing. In the workshop you’ll gain a solid understanding of how these components can work together to help you bring machine learning models to production faster. Click to schedule a workshop for your team.
At Arrikto, we are active members of the Kubeflow community having made significant contributions to the latest 1.4 release. Our projects/products include:
- Kubeflow as a Service is the easiest way to get started with Kubeflow in minutes! It comes with a Free 14-day trial (no credit card required).
- Enterprise Kubeflow (EKF) is a complete machine learning operations platform that simplifies, accelerates, and secures the machine learning model development life cycle with Kubeflow.
- Rok is a data management solution for Kubeflow. Rok’s built-in Kubeflow integration simplifies operations and increases performance, while enabling data versioning, packaging, and secure sharing across teams and cloud boundaries.
- Kale, a workflow tool for Kubeflow, which orchestrates all of Kubeflow’s components seamlessly.