Overview
Reginald Dewil is a co-founder of Penumbra, based in Leuven, a company he started in January 2025 and where he is chief executive. Since August 2023 he has also been a partner at PEOPLE EXPONENTIAL.
Before that he worked as an optimization consultant at Hatamoto from October 2022 to August 2023. He held several academic posts, serving as assistant professor in decision making in Industry 4.0 at KU Leuven from December 2019 to October 2022, and as assistant professor in operations and supply chain management at Vrije Universiteit Amsterdam during 2019. He was a post-doctoral researcher at KU Leuven from May 2017 to December 2018 and again from October 2014 to May 2016, with a period as an algorithm engineer at CONUNDRA in between. He had earlier been a research assistant at KU Leuven from 2009 to 2014, an analyst at Accenture from 2008 to 2009, and a process engineer at imec from 2006 to 2007.
Dewil earned a PhD in engineering from KU Leuven in 2014, following a master in industrial management, transportation and logistics there and a master in industrial sciences, chemistry, at Groep T - Internationale Hogeschool Leuven.
Career history
In the news
- Our Scheduler is getting a major update this month. Scenario Comparison. Side-by-side production schedules highlighting differences in lead time, on-time delivery, and setup time. Easily the most requested feature in every conversation this year. Disruption Modeling. Machine breakdowns, overtime, or a sudden rush order. Simulate the exact cost impact before you make a call. Advanced Constraint Handling. Operator certifications, tooling availability, molds, and sequence-dependent setup rules. That last point is the core of how we
- Why sequential planning almost always costs money. A planner starts with the first coil, finds the best combination of pieces that fits, and moves on to the next. At every individual step he makes a good decision. That is the treacherous part. But coil two inherits the offcuts from coil one. Coil three inherits those of the first two. By coil twenty there is no good decision left. That is called a greedy approach. Locally optimal, globally mediocre. It does not only apply to cutting. Any schedule built up order by order has this
- In 2017, I already knew I wanted to build a planning software company. I didn’t actually start until 2022. What happened in between? A postdoc at KU Leuven, followed by professorships at Vrije Universiteit Amsterdam and KU Leuven. Intellectually, it was the richest environment I had ever worked in. And yet, it was still procrastination. I told myself I was waiting for the right moment. In reality, I chose the safe path simply because it was safe. Academia rewards depth, but not results on the factory floor. And that was exactly
- Your batch sizes are probably wrong. Not because they were calculated incorrectly. Because they were calculated. Someone once made a trade-off between changeover cost and inventory cost. A good trade-off, at the time. Since then the demand profile has shifted. Products have been added. The changeover time on that one machine has come down. Inventory has become more expensive, and the batch size of 400 units is still there. Seven years on. Batch sizes are dynamic, not static. They belong in your optimisation, not in your master
- I ask every operations manager the same question. What makes a schedule good or bad in your plant? I rarely get a clear answer. Usually something like: it just has to work. That is not an objective. That is a constraint. Minimise changeover time? Then you risk late deliveries. Minimise tardiness? Then you change over more often and lose capacity. Minimise lead time? Then inventory drops but you become more vulnerable to disruptions. No schedule wins on all three. Only a schedule that makes the trade-off you want. The problem is
- Onze fabriek is te specifiek voor planningssoftware. Onze planner voelt het beter aan." Ik hoor dit in zowat elk eerste gesprek. De eerste helft klopt meestal. Er zijn mallen die maar op twee machines passen. Er is een operator die als enige de oude lijn kan opstarten. De tweede helft klopt niet. Wat er eigenlijk gezegd wordt: de standaardpakketten die wij bekeken hebben, konden onze werkelijkheid niet aan. There is no free lunch. Een generieke tool moet in duizend bedrijven werken en modelleert daarom geen enkel bedrijf goed. Dat
- Yesterday, around 35% of Earth’s surface experienced the solar eclipse from within the Moon’s penumbra. So, what does that have to do with our startup’s name? A penumbra is a region of partial shadow, between light and darkness. Optimization theory has a similar boundary: between feasible solutions (light) and infeasible ones (darkness). The best schedule, route, plan, staff roster, ... theoretically sits right around that boundary. But real operations rarely have a crisp boundary. Capacities are estimates. Processing times vary.
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.