Episode description

Max Saltonstall and new host Anu Srivastava are in the studio today talking about Vertex Explainable AI with guests Irina Sigler and Ivan Nardini. Vertex Explainable AI was born from a need for developers to better understand how their models determine classifications. Trusting the operation of models for business decision making and easier debugging are two reasons this classification understanding is so important.

Explainable models help developers understand and describe how their trained models are making decisions. Google’s managed service, Vertex Explainable AI, offers Feature Attribution and Example Based Explanations to provide better understanding of model decision making. Irina describes these two services and how each works to foster better decision-making based on AI models. One or both services can be used in every stage of model building and to create a more precise model with better results. Example Based Explanations, Irina tells us, also makes it easier to explain the model to those who may not have strong technical backgrounds.

Ivan runs us through a sample build of a model taking advantage of the Vertex Explainable AI tools. Presets provide easier setup and use as well. We talk more about the benefits of being able to easily explain your models. When decision-makers understand the importance of your AI tool, it’s more likely to be cleared for production, for example. When you understand why your model is making certain choices, you can trust the model’s outcomes as part of your decision-making process.

Irina Sigler is a Product Manager on the Vertex Explainable AI team. Before joining Google, Irina worked at McKinsey and did her Ph.D. in Explainable AI. She graduated from the Freie Universität Berlin and HEC Paris.

Ivan Nardini is a customer engineer specialized in ML and passionate about Developer Advocacy and MLE. He is currently collaborating and enabling Data Science developers and practitioners to define and implement MLOps on Vertex AI. He also leads a worldwide hackathon community initiative and he is an active contributor in Google Cloud.

• Unify data lakes and warehouses with BigLake, now generally available blog
• What it’s like to have a hybrid internship at Google blog

• Vertex AI site
• Explainable AI site
• Vertex Explainable AI docs
• Vertex Explainable AI Notebooks docs
• Feature Attribution docs
• AI Explanations Whitepaper site
• Explainable AI with Google Cloud Vertex AI article
• Why you need to explain machine learning models blog

Anu just got back from a nice vacation and is picking back up on how to use our AI APIs with Serverless workflows. She’s working on some exciting tutorials for our AI backed Translation API.

Max just got back from family dance camp and is working to make excellent intern experiences.

Max Saltonstall and Anu Srivastava