Aspen Technology: Ask an Industrial Data Scientist, with Heiko Claussen

The industrial workforce is undergoing major and rapid change. Two competing pressures are at play to force these changes: widespread digitalization that takes advantage of new technologies such as industrial AI and a growing wave of veteran employees retiring. These two trends have created the opportunity for a new type of role in the process and industrial engineering sector: the Industrial Data Scientist. Bridging the gap between traditional data scientists and domain experts, industrial data scientists bring a fresh perspective to process engineering data management practices.

To get a better insight into the mind and work of Industrial Data Scientists, we reached out to one of our own Industrial Data Scientists – Heiko Claussen, AspenTech’s Senior Vice President of Artificial Intelligence, for an inside perspective on the journey of industrial data scientists and what their role can bring to the table for today’s process engineering and manufacturing companies.

What is the typical “daily” routine of an industrial Data Scientist? Is there even a typical routine?

It really varies, both depending on the company and the Industrial Data Scientist themselves. There are different flavors of industrial data scientists – some are like subject matter experts, others are more like researchers, and others may be closer to data science and coding. It’s not a black-and-white binary of who is or isn’t an industrial data scientist; it is a larger landscape. There are some things Industrial Data Scientists always do – stay up to date on day-to-day workflows and collaborate closely with product management or customers. Continuous learning is also a big part of an industrial data scientist’s job – reading research papers and keeping up to date with what’s new in their field. In this sense, part of their routine is simply to follow what is constantly changing.

How do Industrial Data Scientists work with traditional Data Scientists? Are the two completely separate from each other or is there an overlap?

Industrial Data Scientists are domain experts at heart, with strong technical backgrounds and an understanding of critical technologies such as machine learning algorithms and Industrial IoT assets. They mainly focus on their own area, their corner of the business. Traditional data scientists, on the other hand, focus more on improving toolchains and algorithms across the plant, regardless of specific domain applications. They operate in different corners of the organization, but there is also a lot of natural collaboration. Industrial data scientists leverage the input of traditional data scientists to adopt features that help maximize scalability or business impact. One side naturally helps the other, but it takes leadership to help in this collaboration.

What are the main obstacles you have encountered that can hinder the work of industrial data scientists?

As I mentioned earlier, landscapes are always different – ​​every company is different, every industrial data scientist’s role or background is a little different from another. But there are a few generalities that I have seen repeated many times. One of the biggest is custom toolchains. Toolchains should be reusable and flexible; When customized for specific purposes, they cannot be automated for broader data cleansing, which slows down work and leads to inefficiency.

Another big hurdle is more structural: how and where data scientists, both traditional and industrial, are used. Different data scientists have different areas of expertise. If pure data scientists are deployed in a space that requires domain expertise, that means they need to collaborate with domain experts to do their jobs. But domain experts have their own jobs, which means they have limited bandwidth to collaborate with data scientists. It also slows down the work. Both problems begin and can be solved by leadership. Management needs to ensure that they have cultivated the right environments for data scientists and industrial data scientists to collaborate and support each other, to maximize efficiency and scalability for all.

How do Industrial AI and AIoT fit into the job of Industrial Data Scientist?

They are intertwined. The goal of the Industrial Data Scientist is to create scalable solutions that use cloud and real-time data at the edge. This combination of domain expertise and industry data is needed to support models with maximum business impact. Industrial AI applications and AIoT are key ingredients to achieve this.

How do you see the role of the Industrial Data Scientist evolving? And how will this evolution also help to evolve their businesses?

As industrial data scientists grow in importance in our industry, they will naturally have more points of contact between different critical technologies within the business. For example, digital twins, first principles, sensor-to-edge toolchain optimization, automated operations, industrial AI applications, and MLOps. Combine contextual information from different data sources to run what-if scenarios, producing more accurate information in real time. These are all functions industrial data scientists can serve, or technologies they can leverage, to help drive new efficiencies for their business. These efficiencies also open the business up to new markets that were previously cost prohibitive due to all the inefficiencies we are currently working against.

You mentioned earlier that some of the biggest structural obstacles facing industrial data scientists are problems that can only be solved through leadership. How can leaders best organize their industrial data scientists for success – and how should that success be measured?

It is important to facilitate collaboration, not only between data scientists and industrial data scientists, but also with product managers and customers. Leadership must cultivate an enabling environment. Business leaders should also consider streamlining toolchains, so data scientists don’t have to build custom toolchains every time. Emphasis should be placed on a strategic approach to industrial data management. Data is the memory of an organization; this is what enables search and automation. Leadership needs to move from the old days of mass data collection to a more thoughtful data management strategy. The more leadership can streamline the processes and workflows industrial data scientists work in, the easier it is to turn their findings into tangible business and customer value.

That’s really how you measure the success of an industrial data scientist: maximizing customer value, and how that value depends on the contributions of data scientists. Look at the areas of the business where the emphasis is on research and innovation. How many ideas are created or validated? How quickly are these ideas validated? What interest do they arouse among customers? How many of these ideas become products or product roadmaps? From there, you can track metrics like usage, impact, customer feedback – all of which inform the value you create, and ultimately the role industrial data scientists play in the successful creation of customer value.

Thanks for your time, Heiko!

Want to learn more about Industrial Data Scientists? Connect to our Webinar Rise of the Industrial Data ScientistJanuary 27, for a panel discussion on how industrial data scientists can help improve your business, including: how industrial data is used in key industries; the role that industrial data scientists can play within your organization; and how to hire an industrial data scientist for your team.

Register for the webinar here.

Sean N. Ayres