May 29, 2023

Exploring the Dutch AI Ecosystem with Michel Meulpolder

Repost of an interview by Kai Lemkes in his newsletter 'The Dutch AI Ecosystem'

Kai Lemkes

Founding Partner at Data Science District

Michel Meulpolder

Managing Partner

In this edition of The Dutch AI Ecosystem, I had the privilege of interviewing Michel Meulpolder, Managing Partner at BigData Republic. As part of our new recurring segment, "Kitchen Insight," I will be interviewing more leaders in this field to provide you with valuable insights and perspectives.

How did the market develop and do you see uncertainty in the market at the moment?

I see three important developments. First, a few years ago, most large organizations were focusing on exploring data science use cases and building proof-of-concepts, which was very valuable but was also expensive. Over the last few years however, we have seen many organizations move their focus from innovation to production. We see that by now, especially in the uncertain financial climate, innovation has come under pressure, while it has become more urgent to make existing solutions work and create value. This naturally goes hand in hand with a move in demand from data science expertise to data engineering expertise, which is challenging since good engineers are still not that plenty in the market. A second trend is that tools and platforms are becoming more mature, which can make it easier to develop and deploy data-driven applications under some circumstances. We have seen a strong rise in user friendly, cloud-based machine learning tools and 'one-stop-shop'-vendors. This is beneficial for organizations, but I think that the tension between standardization and customization, which has always been a dynamic in IT, is just as applicable to developing data applications as it is to software engineering in general. In that sense, we rarely see large organizations tackle complex challenges with out-of-the-box solutions. For smaller organizations however, this might be a great starting point. The third development is no surprise: the explosion of AI which gives a new dimension not just to the applications, but even more so to the way that these applications are being designed and implemented. AI can find use case for you, can come up with baseline models and can write or suggest the application code needed to implement them. I think this definitely change the nature of data science work and create a shift from building to integrating. After all, an AI like ChatGPT can do a lot, but it can't integrate itself in an organization.

What kind of projects is BigData Republic involved in?

We mostly work for large organizations in various domains such as energy, retail, telecommunications, transport and public services. Our expertise is usually required in complex projects where we supplement teams at our clients with specific skills, both technical and consultancy wise. We have quite a strong software engineering mindset, and are mostly involved in developing and deploying data intensive applications (such as machine learning applications), and the platforms on which these applications can be brought to production. In the end, this enables use cases such as energy supply and demand prediction, next best action recommendation, logistics optimization, predictive maintenance, financial forecasting and many more. As I mentioned, off-the-shelf products are not feasible for these clients, so they need consultants and engineers to design, deploy and integrate these applications within the context of these specific organizations. I am proud that for many years we have been contributing to tangible solutions in organizations that are crucial to society.

What challenges do our clients have, and which kind of mistakes do we see happening?

First of all, productionizing is still a major challenge. It is one thing to come up with a good model or script, but it is an entirely different ball game to turn this into a mature product, serving the end users in a stable, scalable and secure way, while being integrated into both the technical environment as well as the business processes. It is still often underestimated, since a lot of this work has to happen 'under the hood' and will seem less spectacular from a business perspective. However, this is where the difference between success and failure often resides. "Begin with the end in mind" is a great principle here as well, and we always advice to take production and integration aspects into account already at the very first stages of use case brainstorming. We do see organizations mature in this, but yet it is not at the level where we think it should be. Organizations with multi-disciplinary teams and an iterative, agile way of working generally fare a bit better in this than organizations where the data scientists are in different teams or departments than the data engineers. Lack of a strong vision at management level seems an open door, but is all too often the case. Moreover, the turnover rate of both managers and specialists is high, which makes it even harder to have a solid, consistent vision and execution.

What is important to take into account for organizations that want digital transformation?

To become data driven you need to engineer your company around it, and not just have it 'facilitated' by certain IT services and applications. Digital transformation is of course driven by technological developments. Over the past three decades, organizations have been used to regard everything IT as tools that you use in your business. This old "business versus IT"-way of thinking can be a major obstacle. If organizations are built this way, it takes a very strong and consistent vision in the highest levels of the organization to change this. Yet, on the long term it is the most important one. We have developed an ML strategy framework which consists of five pillars: people, process, organization, technology and data. You need to implement a holistic vision where all of these aspects come together in order to transform successfully. Within this vision, it is important to think in concepts, not in specific tools, products of vendors.

What is your vision for the future, everything seems to revolve around AI nowadays?

Well, as said, no matter how impressive any technology is, it will not integrate itself into our world. Well, maybe one day, but then our view on it probably doesn't matter so much anymore. But as it stands, with all current kinds of AI systems, it is up to us to integrate them into our environment. This will need a specific skillset and way of thinking, which creates a lot of new work which replaces other work. Large organizations might have been ahead in terms of data science for many years, but it is very well possible that they will be caught in the law of diminishing returns if compared to small, agile new companies. New companies will be built from the ground up facilitated by AI technologies. They will not only be "digital first" companies, but "AI first" companies. AI technology will become just as normal in our daily life as other technologies have become over the past decades. Some of it will probably be astounding and (almost) magical. And yet, despite all of this, I myself believe that all of these developments will not significantly change the actual nature of businesses, people and the world. Our tools and circumstances continue to change, but our desires and fears are the same as what they have been for thousands of years.

What do you want to tell the readers? Any takeaways?

Everything is temporary - enjoy it.

Don't miss out on new articles about the Dutch AI Ecosystem!

Subscribe to The Dutch AI Ecosystem Newsletter