Innovation is therefore not necessarily to change and to use new products, but it is to select the best tool to solve the problem you have to solve, trying to focus on the quality of your solution, your models and your ideas.

— Stefano Gualandi

I am a researcher at the Dept. of Mathematics “F. Casorati”, University of Pavia. I’ve been working here for 2 years, starting November 2016. Last year, my department was awarded as one of the Departments of Excellence of Italian universities. The award consists of a five-year (2018-2022) Italian Ministry of Research and University special funding to strengthen and enhance the excellence in research and teaching.

My research activities focus on Computational Methods based on Mathematical Optimization to solve challenging problems. My profile is characterized by a degree in computer engineering, a master in artificial intelligence (in Belgium) and PhD in operational research at the Politecnico di Milano.

After the PhD, I worked for 2 years in Milan and for 3 years in Pavia with a research grant from Regione Lombardia. Then, I spent 3 years in Switzerland (Lugano) working at AntOptima, a spinoff of the Swiss AI Lab IDSIA (Istituto Dalle Molle di Studi sull’Intelligenza Artificiale), directed by Luca Gambardella. This last experience gave me a unique opportunity to work on industrial projects: it was very useful because we didn’t focus directly on basic research, but we developed new industrial products, starting from the latest research results.

 

We interviewed Stefano Gualandi, winner of the Resource Allocation Optimization Mathesia project.

Do you often work on industrial projects?

I started working on industrial projects during my PhD: I had the opportunity to face a problem of optimization of waste collection routes.

After the PhD, during the research grant I worked a lot with a software house, which sells products based on combinatorial optimization methods. We developed new optimization algorithms in order to update their portfolio. One of their customers was the Milan ATM, the company that manages the public transportation in Milan. For ATM, together with Federico Malucelli, from Politecnico di Milano, we worked on two projects: the first one for the staff turnover and the second one for the online management of service disruptions.

What are the challenges of working directly with industrial customers?

In almost all the projects I attended, there were always another company and other people interacting with the end users. This makes work much easier for an academic researcher. In the case of ATM, they were the end user, but we worked with a software house. The software house already knows the customer and his problems, and they have already developed a common language. At the same time, the software house developers have technical-scientific skills and are able to dialogue with us, the mathematicians, promoting the technology transfer.

Mathematicians can develop new models and software to prove that a new idea is valid, but then we need someone else to develop a commercial product and to provide assistance when the product goes into production.

Is working on university projects and on industrial projects complementary, or are they two different things?

It is difficult to give an answer that suits everyone, because it depends on the research topics. There is an overlap, but there are also differences. The first big difference is in the objectives: a company wants to develop new commercial products to sell to their customers to make profit. Moreover, a company usually works with a very short times horizon.

While from an academic point of view, the goal is, or should be, to gather new knowledge. We take the time to look for the “best” solution, while discarding unsuccessfully options: it is the quality of results that matters the most.

How is this dynamic relatable on Mathesia?

The project I won on Mathesia for me was a big success: we made the company happy and we published two research papers (1, 2).

In this project, I developed the software, a dynamic linkable library that the company integrated into its application as a new plugin. They put the new plugin into production, then in the testing phase new requirements were introduced, and, hence, we started a new second project. On both projects, we worked at different levels. The company has an IT department, so there were IT technicians, and it was easy to get the information and the data I needed because they had a technical language. The project manager has a degree in statistics and she interfaced with the company higher level managers.

While working on this project every now and then I consulted with a mathematical physicist – Giuseppe Toscani – I told him that I had observed a certain particularity in their data. Then we asked ourselves “Why does the data have such distributions? How can we explain the formation of this shape?” We started to study deeper the “anomalies”, and in the end, we developed a research project, which culminated in publishing two journal papers, which are not directly related to the company application. However, if I had not worked on that project, then, we have never had the chance of looking at that kind of data.

Hence, working for a company is not a waste of time, rather it is a big opportunity. Many times, there are interesting aspects or a particular subproblem that, if solved, can stimulate new research, and the results can be published in a peer reviewed journal.

Applying for the Mathesia project I would not have been able to foresee that kind of return, but if I had not put myself in the game, surely nothing would have arrived.

How did you get started on Mathesia? What’s the value you find in it?

I discovered Mathesia thanks to word of mouth. Daniele Boffi, from my department in Pavia knew the platform thanks to colleagues from Milan. He sent me an email pointing out an optimization project. I looked at it and tried to write my proposal, taking advantage of the experience I had in Switzerland working at AntOptima: I tried to make a development plan with a project schedule divided into phases.

I really appreciated the tight deadlines of Mathesia: since I made the proposal to when I had the answer the times were quite tight … especially if we compare with the application to the Ministry of Research funding opportunities. Additionally, I appreciate having a project description and a forum: I did not use it, but I read the questions others did.

From my prospective, project budgets can sometimes be limited. An industrial project normally starts with a deep analysis: we must acquire a common language and help the company to understand its problems. And then usually the industry does not want just a detailed report, but a software or at least a software prototype, and this takes time.

What tips would you share in the best practice of submitting a proposal?

First of all, I would advise everyone to try to make proposals, at least to get involved. Because often among colleagues I notice a little fear of being out of balance. I think you have to try, otherwise you always remain closed in your “comfort zone”. It is easy to choose a theoretical, pure problem and work on that. Finding someone who offers you a difficult problem means getting involved, because maybe you will not be able to solve it.

The spinoff experience has helped me so I structured my proposal in different phases. From the management point of view, it is necessary to give a clear planning, making an estimate of the time required for the different tasks. In my opinion, with a good structured plan, it is easier for the company to understand its ROI.

I think that for an academic researcher it is more common to focus on using a certain method or to cite the most recent works. This is certainly important, but structuring well the proposal is more important.

What is innovation to you?

Innovation is a somewhat overrated or misused word. First, in my opinion innovating does not necessarily mean that there must be a change, or that a new tool must be used. In my opinion, the real innovation is using the right tool for the right job.

Second, true innovation is quality work, a job well done. And a job well done implies having chosen the best available tool. For example, if I have to plant a screw in a closet to fix a door, I can take a hammer and give some hammers to the screw. For a few days the door might stay in place, but after a few days… Maybe it’s better if I stop for a moment to think and decide to use the screwdriver.

From the point of view of basic academic research, we must study and invent new tools. However, it is not always true that to solve a given industrial problem, it is necessary to use the latest “trendy” tool or product.

Now everyone is talking about machine learning. Which is perfectly fine, but only if the problem that needs to be solved is a problem in which it is correct to use machine learning techniques. If instead, we recognize that is better to use basic inferential statistics or combinatorial optimization or finite elements … why should we use machine learning?

Innovation is therefore not necessarily to change and to use new products, but it is to select the best tool to solve the problem you have to solve, trying to focus on the quality of your solution, your models and your ideas. But for quality it takes time: it is not a thing that is done in a hurry.

 

References
1 – Gualandi, Stefano & Toscani, Giuseppe. (2018). Call Center Service times are Lognormal a Fokker-Planck Description. Mathematical Models and Methods in Applied Sciences. 28. 10.1142/S0218202518500410.
2 – Gualandi, Stefano & Toscani, Giuseppe (2019). Human Behavior and Lognormal Distribution. A Kinetic Description. Mathematical Models and Methods in Applied Sciences. In press.