SDS 435: Scaling Up Machine Learning — with Erica Greene

Managed Services

Erica Greene, a machine learning manager at Etsy, shares insights into the managed services that power their ML infrastructure. She highlights the use of Google Cloud's managed services, such as AI platform training and hyperparameter tuning, which allow for scalable and efficient model training 1. These services enable teams to run models on various hardware configurations without the need for constant server management 2.

Managed services as like a broad category of these cloud based services where you're not, you know, you're not hosting your database, even if you're sort of renting the server space, you're not running the database.

--- Erica Greene

Jon Krohn and Erica discuss the rapid evolution of tools like TensorFlow, emphasizing the importance of staying updated with version changes to maintain compatibility and functionality 3.

   

Collaboration Tools

Collaboration is key in scaling machine learning efforts at Etsy, and Erica Greene discusses the tools and practices that facilitate teamwork. She mentions the use of Comet for experiment tracking, which has been a game-changer in moving away from cumbersome spreadsheets to more efficient tracking systems 4.

We ended up going with Comet, but I think it's one of those things that zero to anything is like night and day, right?

--- Erica Greene

Remote collaboration strategies have also evolved, with a focus on asynchronous decision-making and minimizing meetings to enhance productivity and team morale 5. Erica emphasizes the importance of shared resources and best practices to ensure seamless collaboration across teams 6.