By HANNAH MARTIN
Hagen Hoferichter is a data scientist, but not just any data scientist. He specializes in the intersection of Machine Learning and embedded devices in the IoT world. Instead of being an obstacle, Hagen’s background as an Electrical Engineer gives him a unique edge when it comes to creating innovative data-gathering devices together with clients. Based on extensive experience working with clients from the healthcare sector, Hagen categorically emphasizes that data science work is more than just programming or number crunching. The most crucial part of his work consists of explaining data science to clients and exchanging expertise to ultimately enable him to create groundbreaking solutions. We met up with Hagen to find out about the most essential ingredients that allows a data scientist to solve intractable client problems.
Hagen, thank you for taking the time to join us today.
Happy to be here.
Let’s start with a bit of scene-setting. Tell us a little bit about how you started working at idalab.
Well, I was actually studying Electrical Engineering at the Technical University in Berlin when I became interested in Artificial Intelligence. I had virtually no background in data science at the time, so I was trying to get my hands on books and articles to teach myself the basics. Then I started searching for companies in Berlin that I could join as a Working Student. Of all the countless options, idalab stood out for me because it was doing very interesting work in multiple sectors, from healthcare and biotech to political forecasting.
So I applied. But..I got rejected.
Yup, I got rejected. It’s funny looking back on it. I’m happy that shortly after I was sent a rejection mail, there was a change of mind. Or actually, maybe a change of heart.
[Hagen gets lost in reflection for a few seconds]
Well, it doesn’t matter, I guess, does it? You’re here now.
Right. I was just thinking about how to distinguish between a change of mind and a change of heart. But anyway! I got invited for an interview, and it went quite well. I was hired as a Working Student, and about a year later, I was promoted to a full-time Data Scientist.
The stars aligned. So maybe you can jump straight into the kind of work you do as a Data Scientist at idalab.
Right now, I’m working on a pretty big project for a client in the healthcare industry. Unfortunately, I can’t name the company due to confidentiality. The client’s main problem is that it has no real way of predicting a patient’s dialysis treatment outcome over time–i.e., whether the patient’s dialysis access will degenerate over time and lead to treatment failure. Basically, dialysis treatment consists of inserting a type of tube into specific veins and arteries from which blood is withdrawn and sent to the dialysis machine for cleaning. Dialysis access refers to the method of insertion and blood withdrawal. For some patients, dialysis access degenerates over time and leads to treatment failure. So obviously, it would be greatly beneficial to be able to predict a specific patient’s dialysis access situation over time, because it would lead to early intervention and replacement of dialysis access with surgery. Ultimately, this increases the patient’s dialysis treatment outcome in the long run.
Long story short, that’s what our client needs, and I’m in charge of creating a solution for them.
So does that mean you collect and analyze data for the client?
No, not yet. The problem is that the client doesn’t have data that would be relevant for us. So we have to create a solution from the ground-up, which means, first of all, that we need to develop a data-gathering solution to begin with. So after a lot of thinking, client meetings, concept prototyping, clinic visits and infrastructure assessment, we came up with a solution that we’ll start testing soon.
Can you tell us what the solution is?
I can’t say a whole lot, but what I can say is that it’s a kind of of audio fingerprinting solution. The idea is to gather audio data from a device, then analyze it to identify relevant audio fingerprints that act as early markers of dialysis access degeneration. And if it works—and we’re confident it will— we’ll be able to identify those patients that show early signs of access failure.
Very interesting. And how do you go about coming up with that idea? What does the process of ideation involve?
That’s the most interesting part of the whole process! Data is only a small piece of the puzzle. Ideation and concept development are far more central to our work, followed by organizational and communication work— that’s where true value creation begins. So most of my work is centered around identifying and talking to the right people in the client’s organization. And, most importantly, getting them together in workshops with the goal of mining expert knowledge and exchanging ideas. Talking to a single expert is one thing, but having multiple experts together is a whole other ballgame. In my experience, that’s when knowledge can be extracted and put to good use.
And so you basically collect expert input, understand the infrastructure, and use them to inform potential solutions?
That’s correct. But of course, the story doesn’t end there. We then enter the world of creating slide decks, which we use to communicate our solutions to clients in a clear and structured way. Actually, you really do two things in a slide deck. First, you synthesize all the expert and other relevant knowledge you exchange from clients. Second, you explain the nature of your solution and why it’s the best possible one given the problem. And the first deck invariably opens more doors, because when clients review it, they realize that some crucial pieces of information were left out during initial discussion. And this obviously enriches our own understanding of the problem further. Another thing is that explaining AI-driven solutions to clients is a delicate matter, and one that needs to be carried out with clarity and purpose in mind. Clear and structured communication offers clients the opportunity to ask further questions and enrich their own understanding of the solution they’re paying for.
Right. So it seems your work involves a great deal of information gathering, synthesizing, and communication.
Exactly. The knowledge cycle with clients is extremely iterative, in the sense that we as data scientists and strategists are tasked with collecting and selecting relevant knowledge first, then synthesizing this knowledge and developing initial ideas for a solution, then getting important client feedback where previously hidden information is revealed, enriching our solutions further, and so on. Structuring and managing this process is really an art that can make or break value creation.
Very insightful. And what role do your colleagues play in all of this? How is it like working with them on creating innovative solutions?
I love working with my teammates. I know it sounds predictable, but I can’t find another way of saying it. People here are extremely intelligent, but more than that, they’re highly empathetic, an important trait when it comes to resolving potential conflicts. Obviously, a few people work on each project together. But the day-to-day work requires a large degree of self-management. Each project member knows exactly what his or her tasks are, and is expected to know how to complete them.
Of course, that doesn’t mean we never ask for help if we find it difficult to do something. The beauty of it is that everyone is eager and ready to help at all times. All they need is a good old calendar invite and help will be on the way. And another very important point is that we’re all from different backgrounds–we have people from engineering like myself, life sciences, linguistics, economics, etc. The availability of different perspectives is of immense benefit to coming up with creative solutions.
We took enough of your time. But one last question. In an ideal world, what kind of project would you love to work on?
To be honest, what I’m doing now with our healthcare client is as close to an ideal project as I can get. First, because of the extent of value that can be created, and second, because the work itself is highly dynamic and multi-faceted. But let me think for a sec (thinks for more than a sec). I’d actually love to work on innovation in the agriculture industry. I can’t think of a specific problem I’d love to solve, but generally I’d love to create solutions that simultaneously drive sustainability and higher crop yields. Yeah, that’s a good way of putting it.
Well, who knows, you might be able to do just that one day.
I’m pretty sure I will at some point.
Ok, thanks for your time Hagen.