#004 - Kubernetes For Humans Podcast with Andy McMahon (NatWest Group)
Sep 13, 2023
auto_awesome
Andy McMahon, a data scientist and machine learning engineer, discusses his transition from data scientist to ML engineer, the challenges of implementing Kubernetes in a large organization, the future of MLOps and Kubernetes, and the enigmatic nature of Kubernetes. He also talks about successful book release and upcoming content on Kubeflow, RNNs, NLP, and generative AI.
The transition from data scientist to ML engineer is crucial, emphasizing the importance of engineering aspects in MLOps.
Successful adoption of platforms like Kubeflow requires critical mass, community involvement, practical examples, and consideration of networking, security, and scalability.
Deep dives
Importance of MLOps in Financial Services
In this podcast episode, Andy McMahon, head of MLOps at a major financial services organization, highlights the significance of MLOps in their operations. The organization handles a massive volume of transactions and data, and McMahon's role is to help operationalize their data science capabilities to create value-added solutions for customers. He discusses the transition from being a data scientist to an ML engineer, emphasizing the importance of engineering aspects. McMahon also recognizes the need for data scientists to embrace the engineering side and highlights the broader industry trend of data scientists becoming ML engineers.
Adoption Challenges of Kubeflow in MLOps
McMahon shares his experience of working in a large organization and trying to promote new technologies like Kubeflow. He emphasizes the importance of gaining critical mass and community involvement for successful adoption. McMahon highlights the need for practical examples and resources to assist engineers and data scientists in embracing Kubeflow. He acknowledges the learning curve and debugging challenges when working with Kubernetes infrastructure. McMahon also mentions the need for organizations to consider networking, security, and scalability aspects when leveraging platforms like Kubeflow.
Benefits of Kubeflow in ML Workflows
McMahon explains the advantages of Kubeflow for ML workflows. He highlights the use of Kubeflow notebooks, which provide a familiar environment for data scientists while leveraging Kubernetes infrastructure. McMahon also mentions the training operators in Kubeflow that simplify scaling models using popular machine learning libraries. He emphasizes the importance of Kubeflow pipelines as a key concept for transitioning from scientists to ML engineers. McMahon underscores the platform independence of Kubeflow due to its open-source nature, which enables organizations to avoid vendor lock-in and explore multi-cloud or on-premise options.
The Future of MLOps and Kubernetes
In terms of future trends, McMahon predicts that large language models (LLMs) and generative AI will play a significant role in the next few years. He highlights the need for upskilling in these areas and the importance of scalable infrastructure, like Kubernetes and Kubeflow, to support LLMs. Additionally, McMahon anticipates increased focus on large-scale infrastructure understanding and building solutions that can handle compute-heavy tasks. He encourages ML engineers to stay humble, recognizing that there is always more to learn and understand.
Andrew P. McMahon is a data scientist and machine learning engineer with several years of experience leading teams that deliver value using cutting-edge technology. He specializes in helping organizations take their initial machine learning and data proof-of-concept solutions through to production. He led machine learning development in companies working across logistics optimization, distributed energy systems, and now in financial services.
He also has a PhD in theoretical condensed matter physics from Imperial College London and has been a part-time science consultant for the Discovery Channel.
His second book on "Machine Learning Engineering with Python" is out now! Get your copy on Amazon: