Statistical information should be part of virtually every business decision, but it isn’t. The reasons why are often more about availability of data and data infrastructure than about the statistical techniques in use. This even though advanced Machine Learning methods can be extremely successful when building scalable models used to challenge established businesses in a competitive market. In this talk, I will discuss our experience at Pace creating a product with predictive models with inductive reasoning at its core; from our early days struggling with data integrity and connectivity to our current period of growth and investment in data science infrastructure and Reinforcement Learning. I will describe the main scientific and technological challenges and takeaways from building a data science product in a product centric company. I will also discuss how we work on model development drawing from common software development practices and how that’s enabling Pace to disrupt the hospitality software market and democratise access to advanced statistical techniques for all hotels.
David is Chief Science Officer at Pace, where he leads the data science efforts and, most prominently, the design of the dynamic pricing engine that is revolutionizing Revenue Management for hotels. Prior to this David was a VP Research at Winton Capital. before Winton he worked on deep learning applied to computer vision problems at Nauto, a Silicon Valley based AI-powered intelligent driver safety company. David holds Computer Science and Physics degrees from MIT and did his PhD in experimental Particle Physics at Caltech. His first exposure to Machine Learning models was using classification models for extraction of the Higgs boson signal as part of the discovery team at CERN. While he considers himself more of a full-stack data scientist, he is currently particularly interested in applications of Reinforcement Learning and decision making in the face of uncertainty.