Search Technology Talk-4

We are excited to announce our first Berlin Search Technology Meetup in 2020.

This time we will have Sylvain Utard & Marwan Burelle (Algolia) coming all the way from Paris to share their insights into search engineering, and Maximilian Werk to give a talk about Zalando's approach to taking data science code into production, also discussing common pitfalls (which will probably ring a bell with many of us).

The meetup will be hosted and sponsored by Zalando – a big thank you!

18:30 - Doors open, Food and Networking

19:00 - Introduction

19:05 - Fast & relevant search: solutions and trade-offs (Sylvain Utard & Marwan Burelle, Algolia)

19:50 - 20:00 break

20:00 - Machine learned search: setting up a production pipeline (Maximilian Werk, Zalando)

20:45 - 21:30 Networking and Drinks

Abstracts

(1) Fast & relevant search: solutions and trade-offs (Sylvain Utard & Marwan Burelle)

Implementing a fast search experience is possible, implementing a relevant search experience is attainable; but having both at the same time is hard. In this presentation, we will explain how Algolia mixes both, what are the performance impacts on some relevancy requirements and what are the trade-offs made.

Bios

Sylvain was Algolia’s first employee and is the VP of engineering at Algolia. Sylvain’s team has grown from 1 to 100+ engineers over the past 6 years. Prior to working at Algolia, Sylvain worked at Exalead and was leading the core-engineering team in charge of the CloudView search-engine and text processing pipeline using C++ and Java. Sylvain is passionate about text mining and web development and along with working at Algolia, he teaches text mining at EPITA, his alma mater. He holds an engineering degree in CS from the same institution.

Marwan is senior software engineer from Algolia’s Search Chapter. Marwan is working on the core of the search engine, with a focus on performances and software improvements in general. Prior to Algolia, Marwan was a lecturer and researcher in EPITA working on various topics such as programming languages and types systems, parallel programming, graph algorithms and malware classification.

(2) Machine learned search: setting up a production pipeline (Maximilian Werk)

At Zalando we have a click-based machine learning engine to improve our search results. In this talk, I’ll present the architecture of this system. The story will include typical pitfalls when productionizing data-science code and our solutions to them. This includes our answers for:

• How to bring a prototype to a production pipeline?
• How to decouple code from the runtime environment?
• Who should to that? Data scientists or DevOps?
• Why are feature flags awesome? And so easy?

Bio

Maximilian has worked as a Research Engineer at Zalando for more than five years. He has worked on automated pricing and full text search. There he has applied methods from Operations Research, Non-linear optimization and NLP while strongly focusing on productionizing data-driven products as well as building fit-for-purpose architectures.