Dr. Marcus Lüdecke

AI (Artificial Intelligence) expert and senior data scientist, Voith Technology & Innovation, Heidenheim (Germany)
Few things, if any, are as rewarding as the inner joy experienced when teams successfully collaborate to tackle a problem.
Dr. Marcus Lüdecke, AI (Artificial Intelligence) expert and senior data scientist, Voith Technology & Innovation, Heidenheim (Germany)

Meet Dr. Marcus Lüdecke, AI (Artificial Intelligence) expert and senior data scientist, Voith Technology & Innovation, Heidenheim (Germany).

Dr. Marcus Lüdecke is a senior data scientist at Voith Technology & Innovation. He studied Mathematics, going on to earn his PhD in Augsburg, Germany. He has been with several cutting-edge-companies over the last two decades, working on various topics such as 2-D-Knappsack optimization problems arising in the context of automated cutting systems, developing medium- to larger-scale databases from a data mining (nowadays also known as AI) perspective, and tackling problems either data-driven or requiring a significant mathematical background.

He has been with Voith since mid-2016. Having started at Voith Turbo, his present occupation is as member of the Artificial Intelligence & Analytics group, which takes on the challenges of producing ever more value from data that moves into reach thanks to the latest advances in the field of AI.

To find out more about Dr. Marcus Lüdecke and his work, read in the full interview down below.

Marcus, you are working with AI and machine learning at Voith. Can you give us an insight into your day-to-day work?

To leverage the full potential coming with available data and advances in computational methods to "dig" that data, experts from various fields must be brought together. This involves more communication than one might assume. Participants must agree on common terminology, approach the problem with an open-minded spirit, embrace other points of view, and talk less and listen more. Only in this way a mutual relationship of trust can grow. And it does not come that easy.

Now why is that so important? You see, we're all humans and all too easily fall prey to the stereotype of shrugging at ideas beyond technical approaches that were established over sometimes decades – that "not invented here" attitude rearing its head.

But AI can offer methods that are complementary, NOT competitive in nature. That simple insight sometimes needs to grow over time.

Apart from this collaboration, data preparation is often more time-intensive than is apparent at a first glance. It is often already at this early stage that a deeper look at the data – re-forming it and excluding certain portions of it, should need be – pays off later when the actual "number crunching" starts. Vagueness, flaws and ambiguities get ever harder to handle the further along the data processing pipeline they show up.

Beyond that preparation, the actual "data science activities" start. However, don't be deceived; this is not the point at which one starts working with that data all alone. With the toolbox of AI methods growing at a rapid pace, one alone can hardly master all available tools at the depth required to tackle the problem at hand. So, regularly I resort to colleagues, discussing details of methods available that can potentially offer aid to a problem's solution. And of course, once first early "results" emerge, I tend to get back to the engineering experts, again aligning our point of view, trying to interpret results in common to ensure we're on the right track.

What keeps this activity continuously interesting is the plain fact that with problems partially recurring and new ones emerging, the only constant across all of them is change. A problem might have been encountered earlier, but still mostly likely its present form is slightly different. While this fact keeps posing ongoing interesting challenges, sometimes reviewing these deviations also yields the key to a "more unified approach" not seen before from which both facets of a problem can benefit. While this latter kind of success surely is a rare bird, it definitely – if caught – is a very satisfying and motivating moment.

Please describe the working atmosphere at Voith.

As stated above, a significant portion of daily activities is devoted to communication. As such, I have experienced Voith’s working atmosphere to be very open, collaborative and helpful. While it seems to be part of the very nature of deadlines being tight and posing challenges, I've always seen people willing to come up with a solution within budget and time constraints.

On the other hand, if requests are beyond scope, one of my learnings over the years is to communicate that using straight talk and “no jive,” working out and naming core obstacles, then finding a compromise satisfactory for all stakeholders. In these situations, I have seen participants being very constructive, not focused on who might have failed where but being forward-focused towards a solution. And like it or not, the need to successfully handle these scenarios is not that seldom as one might wish. While careful planning can avoid pitfalls to a certain extent, the question is not "will a problem show up" but rather "when will it show up."

Dealing successfully with these intricacies is what makes the true difference on the road to success.

What innovative technologies are you dealing with? Can you give us a brief insight here?

Let's start with "explainable AI." While it is nice having a system that predicts early on certain kinds of machine failures, for example, this is at best half of the story, as it will only naturally raise the questions "why will it fail and what action needs to be taken to avoid failure?”

In other words, what we are seeking is an understanding of the root cause. One desires insight into "how" whatever AI apparatus came to its conclusions. Unfortunately, the more elaborate the apparatus (e.g., a convolutional neural network), the more difficult in can become to obtain this insight. A lot of active work is taking place out there right now that deals with how to improve the insight into the reasonings of advanced algorithms.

Another exciting topic is the one called "quantum computing." With theoretical foundations laid more than two decades ago, the recent decade has seen the advent of these machines, now actually existing out there and an ecosystem of related activities developing fast. While significant challenges still remain to be tackled in this area so these machines become common place, the potential exists for definite disruption in some areas of computation.

These computers work fundamentally differently and will be able to compute problems every hardware today currently fails to perform within reasonable time bounds. But this fundamental difference in functionality also implies that programming these machines also looks different, not coding algorithms in the usual sense. This is where things get interesting. One has to reformulate problems such that they fit the nature of the machines and their capabilities.

And last but not least, let me mention the field of activities referred to as "strong AI”. While AI referred to as "weak AI" cannot actually “do new things on its own,” strong AI can propose new approaches to solve problems by deeply interconnecting supposedly unrelated areas, generating abstractions by themselves that then serve as bridges between those areas.

You might remember that very question sometimes encountered at the end of a math or physics exam in school. That "transfer" question to which your first reaction was "now what the heck does this have to do with stuff I learned and prepared for the exam?" This is what strong AI is about – making knowledge and methods applicable to very different problems by identifying a common abstract notion present in them both.

Dr. Marcus Lüdecke

AI offers a broad spectrum of tasks. What skills should AI developers have in order to drive forward the digital transformation around Industry 4.0 at Voith?

"Many are the virtues that can pave the way to success." While hardly anyone excels at all aspects at the same time, I'd say these skills should be listed, albeit only forming the more technical part of the job: a proficiency with the basic tools of the trade, e.g., deeper knowledge of at minimum one programming language that exhibits an AI-affinity, like R or python. This knowledge is best complemented by some expertise regarding basic paradigms of programming, such as design patterns.

Apart from these, some math or physics background is not entirely mandatory, but it can help. With new approaches and algorithms emerging at an ever more rapid pace, it is difficult to keep a firm grasp on each of those concepts' core. It can help to be able to dig into those very details, should need be.

"And that's it – well of course not!" As indicated above, there is a second set of components required to do the job, and while non-technical, is equally important: Communication skills and a mindset open to the views of others.

And finally, strong creativity to reach beyond standard approaches is best paired with the very trait every scientific person should bear: curiosity to explore matters.

Why do you think young people or experienced professionals should pursue a career at Voith?

Simple answer: Because you can write your very own success story right here. Why is that? Because from my experience, your voice will be heard, and your input considered. There is a ubiquity of possibilities to "do your thing." If you are keen on experiencing the ever-daily buzz of a small start-up, this opportunity exists to a certain degree, too, many projects however are – plainly spoken – “larger than you”, e.g. involving a “technology stack unseen behind the scenes”, posing various challenges that cannot be tackled by one alone but rather a team. And often projects relate to products that already have a large installed base thus you cannot approach these with a “do-what-is-needed-on-the-spot” approach. While the solid foundation required is known to inject a certain degree of inertia, there's nothing wrong with this but rather a typical trademark of a larger company. On the other hand, if you have ideas, you will most likely manage to receive funding to dig into those further. As already said, projects crossing your way will have many facets offering a multitude of activities to engage in.

Is there anything else you would like to mention at this point?

While I do not know if I'm still exactly young, at least I can say I'm getting older. So let me tell you a story from my experience through all those years, a story about what matters at the end of the day.

There is a saying that states "there is a magic in every beginning" – and that is true. Especially at the early stages of my career, I experienced that enthusiastic spirit while working on new areas of problems. And while I hope this magic will last a very long time in your case, be aware it most likely will see some diminishment over time. You'll get used to a whole "zoo" of different problems and with that, the initial thrill might become less intense. There is nothing wrong with this; it’s a normal and part of human nature.

But what remains rewarding throughout time is the inner joy you'll experience tackling the next problem and finding a solution to it. Not because people tell you you've done a good job. Not because it is item #101 in your list of challenges tackled. But simply because you know for yourself that you did it right.

Want the opportunity to experience that joy? It is here at Voith.

Competent, motivated, and involved

Did you know our most powerful drive at Voith is the talent of our people? Competent, motivated and involved – it’s the people that drive our innovation and global competitiveness.

Meet our employees