PyData Sofia Meetup December 2019

Welcome to our December PyData Meetup in Sofia!

Christmas is coming soon, so we decided to do our "November" meetup on the 5th of December!

Come join us for two amazing talks, beer and lots of time for nice chats! This time we're hosted by Tide (https://www.tide.co/).

NOTE: This time the talks will be in English.

Speakers:
Hendrik Brackmann (Head of Data Science at Tide Business Banking): Hendrik has built and led data-science teams in various financial institutions within London and Berlin. He is passionate about using machine learning to help businesses make better decisions and how to enable data-scientists to iterate as fast as possible.

Dr. Svetlin Penkov: Svetlin works at the intersection of artificial intelligence and robotics and is interested in extracting programs from large amounts of data. He is the founder of Sciro Research - a team of world class AI & Robotics specialists researching, developing and deploying state of the art AI systems. Before that, he did his PhD at the University of Edinburgh and was a Research Scientist & Tech Lead at five.ai, the biggest autonomous vehicles start up in the UK, where his team worked on learning models from data predicting how vehicles and pedestrians behave in dense city traffic. He has been actively involved in the entire process of bringing state of the art research into production code on the fleet of vehicles operating in London. When he is not around robots he likes to read, build with LEGO and play basketball.

Schedule:

19:15 - 19:55
Neuro-Symbolic Models: Teaching Neural Networks to use Computer Programs
by Svetlin Penkov
About: Deep learning has many limitations such as limited generalisation capabilities, black box nature of the models and considerable computational demands. Classic symbolic AI methods can address all these issues, however they do not scale well when applied to large amounts of raw data. In this talk, I will introduce the idea of hybrid neuro-symbolic models and how they can be applied to deep reinforcement learning. I will present my ICLR 2019 paper and will show you how neural networks can learn to use Python programs in order to solve a task.

20:00 - 20:40
Building infrastructure for Machine Learning
by Hendrik Brackmann
About: We are going to talk about the challenges that we faced when we trained and deployed real time decisioning models for credit and fraud as well as the patterns that we found useful for building these kind of systems.

20:40 - 22:30
Beers and bonding

The event on Facebook: https://www.facebook.com/events/690346894781576/