Speaker interview: Till Rohrmann

We are introducing to you some of the speakers of the next Berlin Buzzwords conference 2015. Therefore, we are interviewing #bbuzz speakers ahead of the conference. See what they responded …

Get to know Till Rohrmann.

Could you briefly introduce yourself?
Hi, I'm Till, a 27 years old software developer. Currently, I'm employed by dataArtisans to work on Apache Flink. I started working on Apache Flink while writing my master's thesis and somehow got stuck with it. At the moment, I spend most of my working time with developing the machine learning library for Flink. If I'm not in the office, you'll most likely find me on one of the numerous beach volleyball courts here in Berlin. I'm also really passionate about riddles. So if you happen to know a good one, then share it with me :-) Maybe I can even challenge you back.

How did you get started with developing software?
I started programming when I was 14. At that time, my older cousin showed me some games he "programmed", or at least I thought that he programmed them. I was kind of envious of him and wanted to do the same. Therefore I bought a C++ for kids book and started hacking. Only later I found out that he wasn't actually programming but using a simple editor. Thus, one could say that I started programming out of a misconception.

What would you like to accomplish by giving this talk?
I want to demonstrate that Apache Flink is a really good platform to build highly scalable machine learning algorithms on top of it. But my talk will also show that a naive implementation of the algorithm seldom achieves good performance and that one has to spend some additional brain cycles to optimize the algorithm with respect to the underlying system.

Which in-depth information is your talk going to provide?
I'll talk about the implementation details of the alternating least squares (ALS) algorithm which is used to factorize matrices. The distributed implementation is mainly network-bound and thus, we can greatly speed it up by decreasing the size of redundant data. Additionally, I'll highlight the improvements we made to Apache Flink in order to support the optimized implementations.

When did you start contributing to Open Source projects?
I started contributing to Stratosphere, the predecessor of Apache Flink, when I wrote my master's thesis in 2013.

Many of the nowadays buzzwordly talks have come from the Apache Software Foundation. What do you think makes Apache projects so successful in particular for communities developing complex software?
The ASF especially offers guidance in form of mentors for young projects. This helps them to learn from the experience of other projects and to avoid common mistakes. Moreover, the ASF allows you to reach out to more people in terms of future contributors as well as media attention.

What was the first Apache project you got in touch with?
It was the Apache HTTP Server which I used to host my first own websites. It was at the time of myspace and GeoCities and my website would have perfectly fitted in there. I'm sincerely sorry for all the eye-pain I caused.

What are the risks of going Apache?
You have to use the Apache hosted JIRA infrastructure. And trust me JIRA can be painfully slow ;-) I'm just kidding. Actually, I think there are not many risks for an Open Source project by becoming part of the ASF. Of course, some of your project's processes might become more complex such as releasing new versions or voting schemes, but in my experience these changes helped us to become a more mature project and grow the community.

Corporate-Design: Extragestaltung, Margarethe Hausstätter
Ilustration: cyan, Berlin