We were pleased to have Raluca Crisan, Co-founder & Chief Technology Officer of Etiq, join us on 24th May 2022, for this ART-AI seminar, entitled ‘On the Path of Building Responsible Machine Learning: ML Testing and Monitoring’.
Many thanks to Raluca Crisan for speaking at this ART-AI seminar and to ART-AI staff member, Juilian Padget, who chaired this seminar.
The recording of the seminar, presentation slides, abstract and bio, can be seen below:
Abstract
Unfortunately a saying in the ML world is; “Let’s put it in production and then see what happens.” What we say is “Please don’t, don’t just put it in production and see what happens!”
There are a lot of issues that can set a system like this up for failure from the get-go, but, these issues CAN be prevented and finding and fixing those issues earlier means the model will perform better, the overall risk is reduced, and you are building a responsible system.
Our concept is that you should have this testing in place from build through to production. From leakage, accuracy, unintended bias through to drift, robustness, explainability, all this needs to be tested and tested thoroughly. In this talk we will cover some sources of errors, approaches to identifying them as soon as possible and some mitigation strategies.
Bio
Raluca has 10+ years of experience in data science working with a variety of clients (UK retailers, banks & telco companies). Raluca’s experience spans managing teams to hands-on data product development. She is Co-founder & CTO of Etiq, an observability platform which supports the entire AI/ML lifecycle with Etiq Test, Monitor, Optimise and Explain, and has a particular specialism in algorithmic bias. Prior to Etiq she was Director – Data Science for Merkle Aquila. She has also built the global analytics team for a mobile marketing company as it was being successfully listed on NASDAQ and has co-founded an ML start-up.
