Machine Learning in Production: What I learned from monitoring 30+ models

apply(conf) - May '22 - 10 minutes

It’s a software monitoring best practice to alert on symptoms, not on causes. “Customer Order Rate dropped to 0” is a great alert: it alerts directly on a bad outcome. For machine learning stacks, this means we should focus monitoring on the output of our models. Data monitoring is also helpful, but should come later in your maturity cycle. In this talk, I will provide practical strategies for prioritizing your monitoring efforts.

Lina Weichbrodt

Machine Learning Lead Engineer


Lina has 10+ years of experience developing scalable machine learning models and running them in production. She currently works as a Machine Learning Lead Engineer at the DKB bank. She previously worked at Zalando developing real-time, deep learning personalization models for more than 32M users.