R & D software engineer and Scrum Master, Laurent Pichon today runs the «Engines & Components» development team, designing and building these fundamental blocks underpinning all the solutions and applications in the Geoconcept range. Here he tells us about his journey, his career, what motivates him and how he sees the future.
When and how did you arrive at GEOCONCEPT?
I arrived here at the end of 1998 after studying at an all-round computer science institute called Enseirb at Bordeaux, and a year of military service, which you still had to do in those days! I worked for about 3 years in a tiny company that was already working on something approaching a geographic information system. I then did a lightening stint in another company where I was not happy at all. That’s when I discovered GEOCONCEPT and its GIS approach. You could say it was the love affair of a lifetime, because I am still here and happy to be so after just over 21 years!
What was your first role, and how has it changed over time?
I started as a straightforward software engineer, and was very quickly appointed manager of everything that concerned the user interface of the Geoconcept application. This is a domain that really interested me, and one I could master. My first mission was therefore to turn the very special dynamic that was there into a concrete reality - something that worked in practical terms, and simplify and refresh the Geoconcept user interface to bring it closer in line with the standards of the most «modern» and «user friendly» applications of that era. This theme is still central to everything we do now.
Once that mission had been accomplished, I expressed a desire to work on more conceptual things, and was appointed architect of Geoconcept solutions. This work is ongoing but not very visible, and absolutely essential for the healthy life of an application. It’s all about looking ahead: we design a general architecture, knowing that not everything will be developed in one go. You have to anticipate ways in which it will evolve, and that requires strategic choices… The software architecture will only last for 6 months. It’s a foundation stone that will bear fruit over 5 to 7 years. We are now on the 4 th generation of architecture for the Geoconcept internal engine, and I was lucky enough to design the last 3 generations.
The 3 rd major mission that was entrusted to me at GEOCONCEPT was to manage a development team for things that are still rather «invisible» in a way: this was the «Engines & Components» team. The role of our team is to supply the building blocks that will then be used across the product range. These are components that enable route calculation, map display, and the route optimization engine…
As manager of a team, are you still developing or does management take up most of your time?
It’s true that my job today involves leading and coordinating a team of about a dozen people, each with a particular expertise. But, happily, I still manage to do quite a bit of development work… This is my passion and I have never wanted to stop doing it. In the technical sector, management functions are not the only way up the ladder. You can choose to rise in terms of expertise, and get really good jobs that are gratifying and recognized, while remaining in the technical camp, as it were. Obviously, there is a human side to this, and it is extremely important: whether we are talking about architecture, the engine or components you never work alone, but are constantly interacting with the other teams in the company.
Over 20 years, which changes have had the greatest impact on your work area?
The move over to SaaS or 100% web architectures was definitely a challenge. When I started, our software was geared to run on heavy-duty workstations. When web solutions appeared on the scene, this led us to slice up these monolithic pieces of software into a series of self-contained «boxes» (the famous engine and components), and to separate these from the graphics side and the user interface so they could be used just as well by a standard piece of software as by our web and cloud applications.
At a purely technical level, the algorithms are not the same as they were 10 or 20 years ago, because research has moved on so much! Now we do things a lot faster. Programming languages are also constantly evolving, some disappear just as quickly as they come into view… fashion also has an impact, and this is why strategic watch is such an integral part of the job of an R&D engineer. It’s essential to keep up to date constantly with anything that can be done better, both at the language level, and also technologies and algorithms, so we can always make the most pertinent choices and offer our customers the most powerful solution possible.
Does GEOCONCEPT develop its own algorithms?
Yes, we create most of our algorithms, but the business of GEOCONCEPT is not to conduct pure research. We have partnerships with research labs that are the real creators of new algorithms. We have to stay at the forefront of what this research world comes up with, spot the algorithms that are interesting for us, understand them, and then use them internally, and sometimes adapt them to our specific needs.
What makes the company GEOCONCEPT so stimulating and captivating for you?
It is, above all, the trust placed in me. I have great freedom of movement, and to act on things, even if, of course, I don’t always do exactly what I want: I have to meet needs as they are presented to me but I have the freedom to find the best way of responding to these needs. Each person in my team has this freedom. That’s what I love about this job: the challenge of finding the best solution and the satisfaction of solving a problem. It’s gratifying. And when one also has a certain liberty so you can do the things you like doing, it’s even more agreeable. This is not the case in every company.
You might think that working for so long with one publisher means always having to do the same thing. My journey proves that this is not at all the case with GEOCONCEPT. Of course I am staying in the domain of geographic information systems, route optimization and geomarketing which are the 3 main strands of our product range. But designing applications that are going to be effective in meeting such a diverse range of client needs remains for me an absolutely fascinating intellectual challenge.
Lastly, GEOCONCEPT is not one of those companies with 1,000 employees. The company is very humane, in the true sense of the word, in terms of its size: everyone knows each other, works together, exchanges news: we keep in constant touch with one another, like an extended family.
Is the geographic information dimension better exploited today than it was at the very beginning in GEOCONCEPT?
Yes it certainly is, as in 1998 when I arrived, we were still very much grappling with the problematics of geography, data input, digitalization, and working to increase accuracy and precision…. GEOCONCEPT was a pioneer in bringing intelligence via the medium of geography, firstly to the geomarketing domain and in the geo-decisional area. We have capitalised on our mastery of geographic information to bring many additional services to companies, typically with our route optimization applications. And as more and more data become available, we continue to develop tools that democratize geographic intelligence and enable our customers to extract ever more value.
What challenges might be round the corner for professional applications such as those you are working on?
It is fashionable to talk about AI and computers that do everything of their own accord. Rather than Artificial Intelligence I prefer to talk about apprenticeship. We are going to see more and more software that learns from its mistakes, that knows how to tell if the result is good from the user’s point of view, and to integrate this information to advantage. Our algorithms currently produce very good technical results. But it can happen that these results are not satisfactory for the end user. For example, in the domain of sectorization: if you have a series of points that you want to regroup to create business sectors based on criteria such as «put the same number of clients in each sector». Our algorithms will do this very well, but the shape of the sectors does not always correspond to what users are expecting intellectually. They will therefore tweak the results to give priority to one criterion or another. This is a typical area where learning – machine learning – can help us understand and integrate what a good business sector shape is, for an end user. Today the right result is conditioned to a large degree by development. Tomorrow’s software will still have criteria to respect as input, but it will be able to compare one result with another and decide which is better from the point of view of ground truth. This adds an extra layer of intelligence.
The challenge for us, as a software publisher, is to upload and analsye all this information in a systematic way, within the limitations of GDPR, and other data regulations. The introduction of a digital adoption platform in the latest versions of our solutions is a first step in this direction. This will enable us to know more about the way in which our clients really use our applications, and from this we can deduce what we have to improve at the level of the user interface. This is not machine learning, but it is a form of learning that brings extra intelligence, and, in the end, allows us to respond better to customer expectations.
What are the challenges you most look forward to taking on in the next 5 years?
For me, there are challenges every day. Always new problems to solve, things we have never come up against before and unique solutions to search for. The biggest challenge, as I was saying, is to create applications that are capable of learning and capitalising on what they do. For example, in route optimization there are lots of exciting things we could do to take ground truth into account more expertly, to reduce the gap between the theoretical optimum and this reality, and suggest results that are more and more realistic and operational. This will require more information gathering, more analysis, more comprehension of the different phenomena that produce an excellent result for the user, and a technical approach that permits the software to constantly self-improve and update.