From AI to impact with ML2Grow

(01-07-2020) AI influences our daily lives, but remains difficult to understand. A conversation with Ghent University spin-off ML2Grow sheds light on the matter.

Artificial Intelligence, Machine Learning ... Not a day goes by that we aren't bombarded with these buzzwords. But where lies the line between science fiction and reality? And especially, how do AI and ML impact companies, and how can these technologies optimise business processes and solve concrete problems? Questions that Joeri Ruyssinck and Joachim van der Herten, respectively CEO and CTO of Ghent University spin-off ML2Grow, are glad to answer. As an AI and ML company they devote themselves to finding the right technology to solve problems that organisations struggle with. Based on data, they try to optimize processes, make predictions ... ‘The moment we create something that changes a business process, it's no longer just AI, but an impact created by ML2Grow.’


‘The moment we create something that changes a business process, it's no longer just AI, but an impact created by ML2Grow.’

Shift & reorientate

Being an entrepreneur means every day is a new quest, as the two founders are by now well aware. ‘Simply setting out on this voyage has caused many things to happen. And in the beginning it is, indeed, chaotic’, Joachim readily admits. In 2017, they started their business adventure with specific technology developed at Ghent University that takes a limited number of data points and uses them to predict harvest volumes in greenhouse farming. After a few months, however, it became clear that their revenue model was unsustainable and their client base limited.


In order to compete with foreign markets, Belgian and Dutch greenhouse farming is evolving towards bigger, high-tech companies. But margins are low and every farmer keeps different information about his production. ‘We had to almost start from scratch every time, harmonising data and identifying important parameters. Farmers do see the added value of being able to predict their harvest, but the investment needed is too high. That’s why this never really took off. It doesn’t mean this technology is any less valuable. It has often proven its utility in settings where little information is available, like in biotech. But in the grand scheme of machine-learning things it is a niche application mainly used by academics.’


‘Simply setting out on this voyage has caused many things to happen. And in the beginning it is, indeed, chaotic.’

ML2Grow 2.0

In the beginning of 2019 a management change took place at the start-up. This change also prompted some profound introspection that showed a new course was needed. Not the most obvious thing to do when you're a rookie who recently took the step from PhD student to entrepreneur. The choices you make impact your business in many ways. At the start Joeri and Joachim’s academic background made them the technical backbone of the company. To this was now added ‘entrepreneurship’.


Making a new start was not an easy choice. ‘We had three options at the time: return to the university, look for a job, or take the plunge and assume control ourselves. We already had a wonderful team working for us that supported us completely, and by then we both felt that the biggest boost we got—and still get—comes from a project going live, from creating an impact with ML2Grow. You get this much less in an academic career. When this opportunity to make a fresh start presented itself, we didn’t hesitate long. But it was still a big, uncertain jump’, says Joachim.


A year and a half later ML2Grow 2.0 was a reality. ‘That’s when we made the transformation from technology provider to problem solver. We did have to reorientate a couple of times during this conversion. At first you think, “We can do anything.” We talked with many companies, they were very interested in our concept, but we should’ve had the reflex more often to go at things from the opposite direction. Now we work differently’, says Joeri. ‘We ask questions ourselves: what problems are there, what processes are you going to change if we present you with the perfect technical solution ...? When the answer is “nothing”, there's really no point in even starting. We built up this knowledge gradually by going through projects from start to end, seeing everything it takes to get something operational. If you don't get to that level, you’re just playing around.’ Joachim adds: ‘At the university you accumulate a lot of technical knowledge, and you assume it’s enough to build up a company. But the execution of a process requires a lot of prior preparation, and this whole process is something you don’t come into contact with as a researcher.’


‘The stages we went through to define ourselves and our work also helped us to present a clearer picture of ourselves to the outside world. Since we’re always dependent on data, it is difficult to deliver ready-made solutions. We had trouble getting this across properly in the beginning. Now we have specialised in setting up short projects where we analyse AI’s impact on companies' business as quickly as possible. We examine companies’ context and data, compare these with similar projects from the past, and use this first analysis to estimate how much AI would cost and what added value it would bring. This lowers the threshold and offers companies a better idea of what we have to offer.’

Once a Ghent University spin-off, always a Ghent University spin-off

Their approach has paid off. ML2Grow was recently nominated ‘AI innovator of the year’ and ‘IT Services Company of the year’ by Data News. ‘It fills you with satisfaction when you offer something and it is appreciated. And we are competitive, even with bigger players. We’ve seen this when competing for public tenders or when organisations contact VLAIO with certain AI-related questions. And we compete not only on price, but also on competencies and content. We feel people trust in our technical abilities—a big thank you to the university—but that alone is not enough to make it.’


‘We remain proud to be a Ghent University spin-off. It’s also nice to maintain our ties with the university in this way. But we are especially keen on being complementary and seizing opportunities to collaborate and to build up new technologies. We do feel a growing distance from the academic world because we simply don’t have the time to keep close tabs on all the new developments. But as a company we still benefit from continuing to think about these things and to question ourselves. This is the added value of being located in the university's science park. We initially set up shop here because of practical considerations. We were still partially working for the university at the time and then of course very handy it’s not to have to commute five kilometres to change desks. But just like many others companies we stuck around, and right now there is no reason to move. It’s a fine location, with a lot of cross-over between with other companies on campus and many possibilities for collaboration. There’s also a pleasant atmosphere, with many students, and the university is close by if you need advice. We regularly refer companies to faculty with specific questions, and vice versa.’

From Terminator to reality check

The three years ML2Grow has been operating make it a young company, but it is not alone: the entire AI sector is still in full development. And while today AI and ML influence a great many things in our daily lives—Netflix’s recommendations, Uber’s prediction of the cost of your ride, etc.—because of their very invisibility these technologies are hard to grasp.


‘Sometimes people’s expectations of AI and ML are very high, but in reality it’s often dull, dry stuff. When your neighbourhood butcher hears the term artificial intelligence, the first thing to pop into his head will typically be Terminator or driverless cars, and he will assume that this technology is not going to help him sell more meat. But he's forgetting that AI can predict where he can most effectively use vouchers. There are many misunderstandings, and sometimes expectations can be futuristic. When these expectations are not satisfied immediately, people are often disappointed, and they fail to understand that there are lower but still valid levels of added value.’


Many actors are currently betting heavily on launching these technologies. ‘There’s a lot of talk about the transformation of the port, the new normal ... but the really important thing is to focus on the advantages of technologies that exist today and that have proven their worth, not on what they might be like five or ten years down the road. This is the added value of a community like AI4Growth, which shows companies what AI is capable of, how others integrated it, what lessons were learned ... The sessions in which concrete use cases are discussed serve as a launch pad to get this matchmaking off the ground.’


'There are many misunderstandings, and sometimes expectations can be futuristic. When these expectations are not satisfied immediately, people are often disappointed, and they fail to understand that there are lower but still valid levels of added value.’

The impact of people, doing your own thing, and a business plan

Looking back on their journey with ML2Grow, what would they do differently? ‘We’ve made plenty of mistakes that I guess other entrepreneurs also make, and we still have a lot of work to do before everything goes as efficiently as possible.’ But if they had to give a Top 3, it wouldn’t take long.


‘It’s important to make your own choices, but also to bounce your ideas off as many different people as possible. There are many doors that a starter can knock on, such as imec.istart, KBC Start it, Voka, Unizo ... and the university too offers a lot of support. But TechTransfer or imec.istart only go that far, and then it’s up to you. You have to take charge, put your product on the market, and fly and navigate yourself. It helps to have a number of helmsmen on board with you during this process, people you trust. Relations may get stormy at times, but they do challenge you to reflect, to put things in perspective ...’ Secondly. ‘Don't put yourself under too much stress. Especially in the beginning we would keep track of what other companies were doing, the reports about deals that had been made, nice projects that were concluded. But what we see is not always an accurate picture: sometimes it’s been embellished, sometimes it's the opposite, and sometimes the truth lies in the middle. So you should definitely do your own thing. Our goal has always been to create as much impact as possible’, says Joeri.


‘And a business plan!’, Joachim concludes. ‘It gives you a blueprint of what you want to accomplish the next year. Even when faced with the unexpected, you always have your business plan to fall back on. It shows you the mechanics of your finances, the investments you planned, the projects that are going to be modified ... and this information helps you to look at alternatives, shift things around ... “In preparing for battle I have always found that plans are useless, but planning is indispensable.” This is the strategy applied by American president Dwight D. Eisenhower when he was Supreme Commander of the allied forces during World War Two, and to this day it stands solid as a rock’, concludes Joachim.



Are you also curious on how Ghent University TechTransfer can help you find opportunities in your research?



Then don't hesitate to contact or take a look at!