Abhijit will discuss how Capital One has developed and deployed MLOps practices at scale across our bank. By overcoming key challenges, building tools, and codifying best practices for building and deploying AI/ML through our broader technology transformation, we are enhancing the quality, collaboration, and reliability of machine learning deployment at enterprise scale across our enterprise. This talk will be high level but will include technical examples and concepts to illustrate how Capital One is extending ML across the enterprise.
Within this presentation Abhijit will offer an overview of the foundational work that has gone into our machine learning infrastructure and use it to offer best practices for other enterprises. This includes thoughts on developing common platform foundations with custom Kubernetes solutions and getting teams on the same stack, as well as focusing on collaboration, bringing down silos, and prioritizing reusable components and frameworks across all ML efforts.
Abhijit’s talk will center on five core best practices for deploying ML Ops across the enterprise, including why what we’re doing at Capital One is unique for a major financial services firm:
1) Getting Your Infrastructure in the Right Place: Best practices based on Capital One’s data transformation, our core cloud infrastructure; CI/CD management; deploying containers; security management; ensuring full visibility across the organization. Focus on staying responsible and well-managed.
2) Being Flexible and Well-Managed: Ability to adapt to different governances (i.e to be able to embed automated checks and QA); adapt processes and platforms based on new governance needs/risk; how Capital One as a regulated company accels at risk management and governance.
3) Scalability: How building custom Kubernetes solutions enabled us to build a powerful ML foundation; ability to automate model monitoring and training, ensuring it’s performing well as you push into production; ability to right-size systems to meet the capabilities and needs of the user.
4) Internal Elevations and Builds: How we incubated local solutions and are scaling them within the enterprise; ability to build on top of open source and customize to our needs.
5) Collaboration: Examples of best practices for aligning ML/AI efforts across the enterprise with reusable components and frameworks