Abstract: 

The term ‘human in the loop' (HITL) isn’t new, yet many data scientists and analytics leaders don't know enough about its critical role in developing scalable machine learning models. Surveys and industry interviews reveal that HITL competency is not yet a focus in data science education.

This is concerning as certain kinds of machine learning model development are nearly impossible without a foundational knowledge of the subject Specifically, it is the management, and even the ethics, of human-in-the-loop projects that many industry leaders are unprepared for. On top of that, the applications for HITL in model development and management continues to grow and change, making it difficult to implement best practices without support from a knowledgeable team.

In this talk, we’ll cover why deep learning needs human expertise to succeed and why blending human intelligence and technology to build scalable models for production is the best approach We’ll also explore where humans can (and should) be used in the model development process and, once deployed, how you can make sure that they are labeling the right things. Finally, we’ll discuss how automation and HITL should coexist in your machine learning operations.

You'll leave this talk knowing how HITL is a vital component in helping not only train, but also sustain, your models in production.

Bio: 

Originally from Cambridge, Matt now helps clients move to a data centric ML approach having worked with clients across autonomous vehicles, green energy and fintech whilst providing meaningful work in the developing world. Away from work Matt has a passion for photography, traveling and unusual cars. In fact his passion for unusual cars bought him to import a Nissan Stagea from Japan to the UK.