ICT with Industry 2025 was intense: ‘No-one showed up during the breaks. We had to knock on doors to convince the teams to take a rest’, case officer Kurt Driessens recalls. Still, the week-long workshop is only one part of ICT with Industry. Thorough preparation sees to it that the teams can make the most of the available time – enlarging the change of follow-up projects during the aftermath or a new case for the next edition.
‘Bonds were forged during the workshop’
In 2024, Associate Professor Artificial Intelligence Kurt Driessens of the Department of Advanced Computing Sciences at Maastricht University took over co-ordination of the bachelor Honours-program KE@Work. He initiated a comparable internship programme for the new bachelor Computer Science, CS@work. Apart from that, he co-ordinated and redesigned the masters research projects to the present format. ‘So when I was asked to help organizing the cases for ICT with Industry, it was right up my alley. I teamed up with my fellow case officer Nicole van der Meulen from Surf to carry out this task.’
The approach for ICT with Industry is of course quite different from the Maastricht internships. Driessens: ‘During KE@work our bachelor students work at real life challenges of companies such as Q-Park, Cummins or DHL. That part is similar to ICT with Industry, as it comprises quite intensive communication with companies about their data sciences challenges. The main difference is that KE@work is about a two-year part-time internship, whereas with ICT with Industry all the action is concentrated in one week.’
Emphasis on AI
Van der Meulen and Driessens explored interesting ideas such as a thematic approach and working with various groups on the same challenge in a competition, comparable to the World Solar Challenge for solar cars. These ideas were abandoned – for now. ING brought in two cases in this year’s edition and Alliander and Contractuo one each. ‘What struck me was the heavy emphasis on AI in all four cases. The two ING cases concentrated on legacy software. Maybe this does not seem AI-related at first glance, but the angle was about Large Language Model (LLM) aided automated software and documentation development to replace legacy software. The problem with legacy software is that it is written in long forgotten computer languages. The documentation is sometimes lacking. The team involved in the case investigated how LLMs could help understand the language and system in order to automatically make documentation about the code. The second case concentrated on translating ancient code into current programming language software.’
The Contractuo case was about automating the tedious parts of tender submittance. Driessens: ‘Researchers and people in the corporate world alike will be familiar with building dossiers to compete for grants and contracts. Contractuo aims to use AI and LLM to take over the boring parts. One of the challenges with this is that all legal requirements should be met, even as these are updated regularly. Would it be feasible to deploy AI for this and to still comply with all rules and regulations?’
The fourth case, aiming at optimized grid use on behalf of Alliander, beforehand seemed to be more mathematical than AI driven. ‘Until it turned out, while the team was under way, that the AI option of Reinforcement Learning offered the most promising direction’, Driessens concludes, emphasizing that the cases were in no way steered towards AI.
No pause
Another thing that stood out for Driessens was how hard the teams have worked all week long. ‘We had scheduled breaks meant for relaxing and networking, but no-one showed up. So many ideas, so little time. We had to knock on doors to get the teams take a rest’, he says laughingly. ‘Apart from that, everyone stayed until late in the evening. The enthusiasm, not in the least from the academic leads, to make a real contribution was overwhelming. This enthusiasm proved to be contagious, as Contractuo has already announced to participate next year with new cases.’ Driessens believes that within the teams lasting bonds have been forged, also between companies and researchers. The case owners were really satisfied with the results.
He himself will also attend next year, but he has switched his role from organizer to member of the steering committee. ‘In hindsight, Nicole and I had an easy ride finding cases. For this year’s team it is tougher to find cases, because the cost for companies has gone up somewhat to compensate for a lower income from sponsors. That is really a pity, as chances are that bringing a case will lead to a very good Return On Investment on several levels. But of course we cannot guarantee success – you never can in R&D, can you?’
Thorough preparation
Rest assured, however, that as little as possible is left to chance. Driessens: ‘Although the workshop takes place in one week in February, the academic leads and the case owners already start coordinating their efforts after the summer to hit the ground running. How much computational power is needed for the case? What data can be shared and used? Can real data and code be used or does the case owner choose to use alternative code? Should the participants sign NDAs? As long as your communication lines are open, a lot can be arranged upfront.’
Apart from that, an info-session was held before the start. ‘The participants receive a heads-up about details of the workshop and the case’, Driessens tells. ‘The team members get to know each other and it is assessed what knowledge each team member brings in to be able to make optimal use of that expertise. And it also helps to strengthen the participants’ involvement. A week before the workshop everything must be clear and ready: the team, its capabilities and assignment and the facilities.’
Expect the unexpected
Asked for things that struck him as remarkable, Driessens says: ‘One of the research applicants was an associate professor. She was attracted by the interesting case. And while she brought in her own expertise in language and the legality of tenders, she seized the opportunity to learn about Large Language Models.’
It worked out for her as it did for the case owners. Driessens: ‘When the week is concluded, it is hardly the end of it. During the aftermath the case owners can draw up the balance sheet: what do the results mean for us and how do we follow up on them?’ Follow up projects with (part of) the team can for instance be brought to the table – as in the case of Alliander. Or, alternatively, thinking about next year’s case can be started – as in the case of Contractuo.
Source Photo: Website Maastricht University