Abstract: ML Governance is often discussed either in abstract terms without practical details. This talk will focus on the day-to-day realities of ML Governance. How much documentation is appropriate? Should you have manual sign-offs? If so, when and who should perform them? Most importantly, what is the point of all this governance and how much is too much?
We'll see that the core of ML governance is ensuring the right kind of risk trade-off decisions are owned explicitly by the most appropriate roles. Many ML governance failures arise from overlooking risks and failing to take risk mitigation steps. We'll walk through a template process that can be used to manage risk and reinforce best practice.
From this talk you will learn:
- What ML Governance is meant to achieve
- How to get started with a template process
- The role of documentation (and especially Google Model Cards)
- Which roles have what responsibilities
- The relevance of a governance board
Bio: Ryan Dawson is a technologist passionate about data. Ryan works with clients on large-scale data and AI initiatives, helping organizations get more value from data. His work includes strategies to productionize machine learning, organizing the way data is captured and shared, selecting the right data technologies and optimal team structures, as well as writing the code to make it happen. He has over 15 years of experience and, as well as many widely read articles about MLOps, software design, and delivery. is author of the Thoughtworks Guide to Evaluating MLOps Platforms.