Overview
Tom Staessens is a co-founder of Leap Technologies, a role he has held since August 2022. He is based in Gent.
From September 2018 to March 2023 he was a researcher and PhD candidate at Ghent University, and during the first two years of that period he also served as treasurer of Engage (Engineering Alumni Ghent). Earlier he completed an internship at Vintecc in the summer of 2018, an internship at JECRC University from July to September 2017, and a student job at LVD in August and September 2016.
Staessens studied electromechanical engineering at Ghent University, taking a bachelor's degree between 2012 and 2016 and a master's degree in control engineering and automation in 2017 and 2018. In 2016 he also obtained a master's degree in the same field at the Universitat Politècnica de València.
Career history
In the news
- One giant leap for ... No okay, we're not quite up and running in space yet. But it's been an honor helping build the groundwork for space-grown food. Excited for what's next.
- The real AI advantage in industry isn't always the algorithm: it's often the people who know the process better than anyone else. Our co-founder Tom Staessens shared his take with Agoria on the future of industrial AI in Belgium. Check it out below👇
- Thanks Agoria for the invitation to share my thoughts on the future of AI in Belgium. My personal takeaway from this fragment: the people in the industry with years of process knowledge are untapped gold for industrial AI applications. As we often say when starting a project: if we don't sit together with your team on what the data means, we'll happily write the walls full of conclusions your people probably already knew. In practice, that domain knowledge pays off best when it meets people who also have a sense of what AI can and
- The last slide claims that intuition on machine tuning is usually more capturable than it looks. Let's be more precise about that, because it looks quite different from one machine to the next. Sometimes the decision reduces to a set of rules. When this condition shows up, take that action. That works mostly when the conditions involved are limited, and once the rules are written down they can be tested and argued about instead of trusted. Often it is closer to a lookup. Years of human experience arise from having seen this
- In Belgium, we won't build the next ChatGPT. For that, we miss the infrastructure and the capital. What we do have, is a combination of world-class companies 'on the size of a handkerchief' as we would say in Dutch. The data that's required to apply AI in the industry does not exist online: it lives in the machines and production lines of those companies. If we capture that data and transform it into applications, we have the resource to become pioneers in 'physical AI': where AI has real impact on machines. Belgian companies are
- A warm start from historical data sounds like a lookup: find the job you ran before and reuse its settings. In practice, the jobs that could use it most are often the ones that aren't in there. New variant, new material, a spec combination nobody has run yet. But warm starting often goes one layer deeper: what matters most often is not the exact same specs, but how the job behaves. Metal springing back differently after bending. Natural fibers twisting and bending depending on their speed. An experienced operator or engineer
- Something we found interesting on a controller tuning project's structure: each stage tends to be useful on its own, not just if the automation works. The first real step is defining the cost function, which is where a lot of the thinking happens. You're essentially forcing the question of what good performance means for this specific machine, in a form precise enough to score. Engineers or operators that have been tuning the same machine for years sometimes find they don't fully agree on the answer. Working through that is
Related profiles
This page shows public professional information only, each fact cited. Is this you? send a correction, or ask for removal within 24 hours, no questions asked.