Luca Falsina

Software Developer, Booking.com

Session

Worldwide machine learning with Bitterballen

How come that I am putting in the same sentence one of the most glamourous (and arguably useful nowadays) fields in computer science together with one of the most glorious dish of the Dutch cuisine?
Let me depict you this scenario: Let’s say you are a data scientist working for a worldwide known company whose mission is to delight its customers by providing them with the right delicacy of the Dutch cuisine at the right time. Although it may seem easy, engineering the best features for predicting such a phenomenon is a difficult task. On top of that, understanding where to retrieve data from, ensure its correctness, and discover what has been already implemented by your fellow data scientists becomes a real nightmare when you have millions of customers using your product every day.
At Booking.com our datascientists face these challenges daily: As developers, our task is to provide them with the right tools and infrastructure to simplify their job and minimize the time frame required before they can actually test their hypothesis in production.
In this talk I will first introduce you to some of these recurring issues. Then, I will specifically target one of the projects we are working on to remove those challenges, the Feature Store, a data pipeline whose goal is to transform basic input collected on our frontend into complex, reusable, and easily-discoverable features, which can be plugged to train machine learning models. Specifically, I will focus on the requirements, and the architecture of such a system, as well as the technology stack and the lessons we learned so far during the development of this project.
Everyone is welcome.. even if you don’t know what a Bitterballen looks like 🙂

Level: Intermediate

Biography

Luca is a Software Developer at Booking.com in Amsterdam. He is currently working to leverage infrastructure and improve tooling to help Booking.com internal teams in taking advanage of machine learning techniques at scale.
Prior to Booking.com, Luca had a six months internship in Amazon in the localization team and before that he studied Computer Science in Milan. During that time he developed Grab’n Run, a library to secure and simplify dynamic code loading in Android apps.
Luca is an eager learner, who enjoys attend conferences and occasionally give talks as well (the latest one were at Droidcon IT 2015, ACSAC 2015, and Droidcon IT 2016).