From the Research Lab to a DeepTech Software Editor
Can you tell us about your academic background?
F.V.: « My opening to the world of business dates back from my year as a fellow at the Sloan School of Management of the Massachusetts Institute of Technology (MIT) in Boston. This gap year came after my Engineering degree in Applied Mathematics from the University of Louvain (UCL). I ended up graduating from MIT, with a second Master degree in Operations Research in 1993, followed by a PhD degree in Applied Mathematics from the University of Louvain (UCL) in 1994.
I started my career as a lecturer at the University of Cambridge in the Engineering department and the Judge Institute of Management Sciences. In 1998, I moved to the University of Bordeaux on a professorship and set up a team in Mathematical Optimization within the CNRS (French National Centre for Scientific Research) research lab.
In 2008, I created a research team at Inria (the National Institute for Research in Digital Science and Technology). The project combined the forces of the Institute for Mathematics and the CNRS computer sciences labs of Bordeaux, along with partnerships with research teams in Rio de Janeiro, Brazil. This project, named RealOpt, developed generic optimization methodologies and their efficient implementation. »
Why did you decide to launch your startup?
F.V.: “Through industrial collaborations in various business areas, our research work has always been challenged by real-world problems ranging from production planning and manufacturing, to telecom network design and logistics. We observed the gap between what academic research is able to deliver and what is currently in use in the industry. Many use cases remain without a solution for lack of appropriate technological tools; and many resources are wasted in the process. Thus, the decision to create our startup was both market-driven and responding to a societal motivation to do more and better, with less resources.
Aiming to have an impact is in the DNA of academics who are trained to publish their results to give value to their research effort. But with DeepTech, making the results known is not enough to achieve the goal of having an impact. Industrial collaborations can bring prototyping, but rarely turn into operational tools. Hence we developed the feeling that it was also our responsibility to carry on the effort up to transferring the technology to the industry. What a great satisfaction to make our 25 years of research and know-how useful for society!
The kick-off for launching this great challenge of industrializing a complex technology was to gather highly motivated and brilliant teammates from our research group as co-founders and a precious external co-founder with business expertise. This adventure is that of a team, in the same line as the team work that drove our Inria project. So the disruptive technologies that had reached maturity were to become optimization intelligence tools in use in business processes via our spin-off.”
What are the challenges of this transition from research to a company?
F.V.: “The fundamental challenge is to make the DeepTech accessible to the user. Atoptima’s raison d’être is to deliver very sophisticated cutting-edge optimization techniques in a completely fluid, reliable and integrated way, with very responsive delivery time. Dealing with an industrial application in an academic context, in comparison, is a different perspective: a one-off, long-run study, only prototyping a solution, with no-integration required.
Every single client’s optimization problem is different. Yet to deliver high performance that scales up, the optimization engine needs to fully exploit the specific structure of the problem. To solve this dilemma, while nurturing accessibility to the technology, Atoptima is delivering custom-made solutions with a customer experience equivalent to integrating an off-the-shelf solver. The machinery under the hood is a very modular software library that encapsulates optimization models and best practices to solve them.
Thus, our internal challenge is to industrialize our know-how in dealing with complex operations planning problems into a generic toolbox of registered, certified and multi-application software components in order to be able to promptly create a tailored application solver for every use case in our scope, and so to make optimization intelligence readily accessible to companies who wish to climb the next step of the technological ladder with easy-to-use cloud-based and dev-free application solvers.”
Do you still feel the vibe of a researcher?
F.V.: “More than ever. Being confronted to real-life applications is the trigger for our continuous research effort to bring better performance, to make our algorithms more widely applicable, and to extend the scope of decisions that can be optimized simultaneously. There are indeed significant improvement margins to achieve by considering, within the same optimization problem, decisions that were previously made independently or hierarchically.
At Atoptima, the R&D effort is permanent and massive to provide unmatched performance on the industry’s most complex operations planning problems, and to contribute to international cutting-edge research in Combinatorial Optimization. Our DNA is this strong scientific background. Our mission is to bring the state-of-the-art via continuous technological transfer. Our recruitment is largely composed of top-notch talents able to deliver this disruptive technology.
Beyond our permanent staff, we continuously develop academic partnerships with best internal researchers in our area. We nurture these R&D collaborations and internal PhD theses. We publish pioneering papers in major international academic journals and conferences. We also contribute to the community via our open source platform (Coluna.jl) on decomposition techniques for optimization problems.”