Thierry Labro: Could you tell us what you've been up to over the last four or five years?

: In 2017, we launched this SnT spin-off. We spent the first four years doing reference projects, to show that the technology had value for different industries. At the end of 2020, we received investment from Enovos and Paul Wurth, to help us make products. We were very focused on specific development services, and rightly so, but we said to ourselves that we wanted to make products to move towards recurrence and licensing. At the time of the investment, there were eight of us. The investment was €1.3m, and we also obtained the Young Innovative Company grant from the French Ministry of the Economy, which brings us to where we are today: we have ready-made products that we are bringing to market.

Alva is the digital twin of electricity distribution grids, with a partnership that we entered into with a Slovenian company, Iskraemeco, which manufactures smart meters. They sell smart meters and will supply the necessary data. Alva will process this data and help create a digital twin to optimise electricity distribution operations. Right now, in September, we are preparing a similar launch for Greycat, our in-house technology, developed 100% in Luxembourg, with no dependence on any other software, which also makes it possible to address issues of sovereignty... To sell this to software integrators, developers... The idea is to say that among all the technologies that exist for data processing and data storage, we are proposing another for making digital twins, which can be augmented with machine learning to create artificial intelligence or for reporting and KPIs.

We're now focusing on commercial deployment beyond Luxembourg. For Greycat, we're going to concentrate on our neighbours and Europe; for Alva, it's directly a global market.

Let's take a step back. Alva, Greycat: just tell us what these technologies can do.

Alva is interested in electricity distribution issues. Our customers are electricity distributors. In Luxembourg, this is Creos for a large part of the country. Creos needs more and more support to manage the network, because of the increase in energy injection by solar panels and wind turbines, as well as the growing consumption of electric vehicles and heat pumps as a result of the energy transition.

This raises two main issues: the size of the grid - the sizing of the cables - which requires knowing where and how much to invest to ensure optimum distribution and control. There was an example, not very long ago: a few houses were disconnected in the North because it was an overcast day with wind and the wind turbines were producing a lot. The grid was stable and at one point there was a hole in the clouds. It passed over a field, the solar panels started to inject power and there wasn't enough consumption. The grid disconnected for safety. Typically, it's on aspects like these that having a digital twin, machine learning and AI will make it possible to take these meteorological aspects into account more finely, more precisely, to help guarantee the distribution of electricity and anticipate these events.

People don't necessarily need AI, there are plenty of industries that are just in the process of digitising, wondering how to collect and store data, and how to analyse it.

Grégory Nain, co-founder and technical director, DataThings

Does it tell us whether the network is correctly sized or not?

For industrial customers, Creos already deploys very specific solutions according to their needs. Luxembourg produces very little. With the 380kw project, they are looking almost 50 years ahead.

What are the difficulties?

The difficulties stem from the sheer volume of data. There are 50bn data points a year in Luxembourg and Luxembourg only has 350,000 smart meters. Strasbourg has 600,000. France has 35m. Everything is connected, and if we want to have very fine granularity in the system, we have to carry out analyses for each distribution point. We learn from each consumption point and each distribution point. We factor in environmental elements and maintenance aspects - from time to time, equipment is taken offline because it has to be maintained - and all this becomes complicated to manage manually. These digital twins make it possible to take all this into account and provide a vision for operators so that they can make investment or operating decisions.

Doesn't the fact that the weather is changing make your models obsolete?

Things change and machine learning can only reproduce what it has seen in the past. We advocate taking this base from the past - because that's what we know - qualifying this data and making predictions, but don't forget to re-train, update and factor at a certain frequency the new data that arrives every month, every three weeks, whatever. You have to enter the new data as you go along.

Isn't there a competitor?

We've found two. Either they can't manage the same volume of data in the same detail as we do in Luxembourg - after 30,000 smart meters, it's complicated for them - or big companies are putting this in their catalogue but they don't have anything tangible. We're ahead of the game, thanks to Creos' vision of ten years ago.

Where do you stand financially with your developments?

We're running at around €800,000 year. We're coming onto the market with a technology that's frightening - people aren't familiar with it. It's reassuring to have competition. Some of our customers have four different systems.

How do you address this global market?

A lot of events, Paul Wurth's network and word of mouth. As soon as we have deployed our solution with one or two customers, things will speed up. The pilot phase lasts three to four months. We take the solution and try to connect it to theirs. In six months, it will be deployed. It's an annual licence that currently depends on the number of smart meters, and that's what's going to determine performance.

What is the ultimate goal of a global project like yours?

Above all, we want to create things that will last. We're not in the startup business, where you burn through cash fairly quickly. In seven years, we've only done one round. We're thinking about another round to support commercial development. We want to showcase these technologies and their potential. It's likely that a major software publisher will be interested. We've given ourselves another five to seven years... The sales cycles for Alva are also fairly long, because we're talking about infrastructure that is going to be national, calls for tender, you have to get them interested...

If we went to France or Germany, we could cut costs by almost a third and still have the right skills.

Grégory Nain, co-founder and technical director, DataThings

With the acceleration of artificial intelligence today, aren't you in danger of competitors catching up?

Perhaps. I don't think that's a risk. You need access to data. We learn from data. In Europe, we're seeing all kinds of regulations - the Act, the AI Act, the Data Act, and so on. Our big competitive advantage is Greycat, which enables us to simplify development enormously. There is no competitor on the market for such an integrated solution. Without integration, performance won't be there. Manufacturers who put our solutions into production are not going to pull the plug tomorrow...

How many of you are there today?

There are 22 of us. A third PhDs, another third masters and a third bachelors, a quarter of whom are in business development. The four founders [Editor’s note: Grégory Nain, chief operating officer; Assaad Moawad, head of machine learning; Thomas Hartmann, head of R&I; François Fouquet, head of software engineering] are still here! For seven years! Plus five years at university before that. We've always had the same vision. We wanted to have an effective technology for digital twins. We see others, for example with Cebi, in industry 4.0, with a digital twin for production lines; in the banking sector, on data preparation... We are beginning to realise our vision. Greycat is useful wherever you want to carry out status analyses. System states. The problem is not analysing data, but the state of a production at a given moment, the state of production an hour ago and the state of production in six hours' time. These states are always elements connected to each other, relational data. Greycat makes it possible to look at these states at a given moment in order to learn from the past and predict the future. Let's try to move our economies from a reactive model to a proactive model!

Or to buy more material if we run out, or to schedule maintenance?

To adapt your business immediately and reach an objective within a week.

This is innovative in the industry?

It's becoming possible with the technologies we have. ArcelorMittal has on spot anomaly detection. It helps them enormously: if they see that the product is deviating, they stop immediately and rectify it. It's almost like immediate quality control. We've been working with Urbasense for several years. They make sprinkler systems with sensors that they place in the ground and controllable solenoid valves. We take care of the back-end: they collect data from the trees in a city so that they can survive. With this, we have models that monitor the root system to prevent the tree from falling with the wind and that will measure how much water the tree needs to survive, so that it limits the heat island effects in summer and so that it doesn't have too much water and doesn't develop its roots.

And in a group with several factories all over the world, will the manufacturer be able to have comparable metrics?

There's that, but there's also... There's a group we work with, Cebi International, which has a plant in Luxembourg and ten others around the world. The aim is to deploy a digital twin in each of the factories and at some point to have a consolidated dashboard of the factories that is permanently up to date. With a sales target, for example, we can have an AI that will enable us to plan our production, that will say ‘plan this production before that one to reduce the changeover time and that one will be more valuable’. A human brain won't necessarily think of everything. But people don't necessarily need AI, there are plenty of industries that are just in the digitalisation phase, that are wondering how to collect and store data, how to analyse it... But it has to support decision-making. If you decide to connect it automatically, you need to put safeguards in place.

With commercial development on a global scale, it's not hard to imagine you moving on from Paul Wurth's incubator?

We're mainly going to be thinking about opening offices in France, Germany and perhaps Belgium, which will be more commercial in nature, so that we can expand. Maybe some technology offices too, because in Luxembourg we're finding it harder and harder to recruit. We're not yet break-even, but it's harder to recruit good people. There are a lot of data scientist profiles that I call 'YouTube data scientists', who have spent three weeks listening to stuff... DataThings is still a very technological company. We have redesigned our programming language, we have our database management system, we have neural networks on GPUs. Most of the people who come to us are technology integrators. Finding the profile we need is very complicated. In the Luxembourg market, it's even more complicated for economic reasons. If we went to France or Germany, we could cut costs by almost a third and still have the right skills. We recruit them in Portugal, Italy or elsewhere, because they are willing to start on more reasonable salaries.

More than ten years is a long time to be working on this project, isn't it?

It's a real marathon. One day it's fine. The next day, there's one thing to sort out, then three or four in quick succession... There are successes and failures, but you're always moving forward. The company is growing, perhaps not very quickly, but solidly. The people are close-knit and get on well. Our production tools are our people. We'd like to reward them, but not just yet. But soon. We hope so. The founders are still here. And some of our employees who have believed in the company from the start. The products are well designed, we can see the markets, we can see the figures. If we sell Alva twice, we become profitable. A second round would enable us to accelerate our commercial development. For example, we're going to do a Greycat tour in France to present our product to both academia and industry.

This article was written for the supplement of the edition of , published on 19 June. Read the original French version of this interview . .