Presented By O’Reilly and Cloudera
Make Data Work
21–22 May 2018: Training
22–24 May 2018: Tutorials & Conference
London, UK

Interpretable AI: Can we trust machine learning?

Konstantinos Georgatzis (QuantumBlack), Martha Imprialou (QuantumBlack)
12:0512:45 Wednesday, 23 May 2018
Data science and machine learning, Expo Hall
Location: Expo Hall Level: Intermediate

Who is this presentation for?

  • C-level executives, data practitioners, data scientists, and machine learning engineers

What you'll learn

  • Learn how to create models to support real-life decision making
  • Understand the differences between interpretability and performance
  • Explore the applications of and use cases for model interpretability


In many organizations today, business decisions are guided by advanced analytics, and concrete actions are driven by predictions made by machine learning systems.

Konstantinos Georgatzis and Martha Imprialou explain how to interpret the predictions given by your black-box model and how machine learning is helping to drive decision making today.

Topics include:

  • Can we trust black-box models?
  • How to explain predictions made by them
  • Interpretability versus performance: An ongoing debate
  • Applications of and use cases for model interpretability: How advanced analytics in industries such as pharma can transform and impact R&D, operational planning, and commercial strategy

Konstantinos Georgatzis


Konstantinos Georgatzis is a data scientist in client projects at QuantumBlack, where he leads the development of methodologies to optimize commercial performance for pharma and healthcare clients, focusing on clinical trial and real-world patient outcomes. He is also developing ways to further improve algorithms that are frequently used within QuantumBlack. Konstantinos has extensive experience in developing machine learning methods, especially for healthcare analytics using biomedical and clinical data. Konstantinos has published eight articles in international machine learning and medical conferences, four of them as first author, and presented in multiple conferences. He also holds a position as a reviewer for international machine learning conferences. He holds a PhD from the University of Edinburgh, where his research focused on developing novel machine learning methods and applying them to better model the biosignals of intensive care unit patients.

Martha Imprialou


Martha Imprialou is a data scientist and a specialist consultant in client projects at QuantumBlack, where she leads analytics work, designs and develops machine learning and statistical models, and works with pharma and healthcare clients, focusing on real-world patient outcomes to optimize commercial performance. Martha has extensive experience in healthcare analytics using genomics and biomedical and clinical data. Previously, Martha was a postdoctoral researcher at Imperial College London, where her research involved analyzing DNA sequencing data to understand the genetics of autoimmune diseases. She has published in five genetics and computer science journals, four of them as first author, and presented at 10+ conferences.

Comments on this page are now closed.


Picture of Mark Atterbury
26/05/2018 10:58 BST

Are the slides from this session going to be shared? Thanks, Mark