 Good afternoon, everyone. So I'm going to talk about a project that we've conducted together with the World Bank. And so the project is really about leveraging any kind of freely accessible data, geospatial information coming from GIS databases, but also low-resolution satellite imagery to make assessments of road network infrastructure at scale. And by that, it means evaluating the condition of roads in developing countries to better plan investment programs and make road networks more resilient to climate change. So first, an introduction for people who do not necessarily know who the World Bank is. It's actually the largest lender in the world. They loan about $70 billion a year for any kind of education, health, food, or infrastructure programs. And infrastructure actually represents 25% of the money that the World Bank is even investing every year. And the World Bank mainly invests in developing countries. And so the context of the work we've been doing at the World Bank, and by the way, we are Altea, we're a software company. We specialize in leveraging any form of data source, mostly coming from visual sensors, to make predictions and work on analysis at scale. And so the context was to help World Bank understand how much and how to prioritize their lending programs tied to road improvements in developing countries. And the main issue that was faced by the organization, it's something that's really simple. The cost of getting access to data in order to make those assessments can be really high. If you talk about high resolution satellite images, it could be more than $1.5 to $2 million just for the data collection process. And so we've worked together trying to think of a method that would actually make the right amount that will give actually the right amount of information while minimizing the data cost. And so the state of the art for geospatial data analysis at scale, if you talk about infrastructure projects, but also, I guess, any type of business topic, you have available open source data that's coming from many different databases. OpenStreetMap is one of them. And that kind of open source data gives you access to, I would say, a first scope of information. In our case, it was the mapping of the road network for the developing countries that we work on. Then you have also low resolution satellite images that are available through different kind of databases. OpenData by AWS is one of them. Or you can also get them directly from the Copernicus portal. And the resolution of that type of imagery is about 10 meters per pixel, which, when it's combined to open source data, can give you great information on the type of the road, if it's paved or unpaved, and the general condition of it, especially if you do not only use optical data, but also SAR information coming from radar satellites. And then on the right side of the screen, what's more expensive is high resolution imagery that will give you more granular information. And usually, you would leverage high resolution images to make like a real microscopic assessment on a portion of the network. So this was kind of like the framework of our analysis. And what we've decided to do was to build an artificial intelligence model to improve artificially the granularity of the low resolution satellite image by multiplying the different data sources that we had access to. So you can see it a little bit as of aggregating so much information together that at the end, the results is a bit bigger than the sum of all the elements that you put into play. And so to do so, we leveraged, as I said, optical data from Sentinel-2, SAR data from the same program, and about a dozen open source databases, GIS, like OpenStreetMap, but also initiatives like Mapillary, which is a way to get access to ground-based images in developing countries. If I take a little bit more, if I talk a little bit more about the approach and get into the technical workflow that we've enabled, it was kind of like three steps. The first step was to build a digital representation of an entire-road network, so at the country level. In our case, we did it for the countries of Peru and Mexico. So linking our AI model to all of the accessible databases that are available in the cloud and work on some aggregation algorithm to build a network model that would be the best representation of that road network. And then what we've done is enrich that representation, the digital twin of our road model network, by getting from the cloud all the optical and SAR information that we could in order to build a pavement index and a condition index. So basically, the pavement index would give you information on if the road is paved or unpaved, and the quality index would be more like how practical is the road. Is it like in a very good state, or is it like completely broke? And that information is crucial because it helps governments, and in that case, the World Bank lending the money. Understand how to develop a specific area. You can understand that if you want to develop a commercial relationship between two cities in Peru, you have to make sure that the highway that links those two cities is actually in great shape and allows the commercial traffic to be handled. So that was the second step. So using that digital twin and reach it with aerial images coming from freely accessible satellite data. And the third step was to do some ground truthing. So using available data that is very granular and available at the country level to understand how good our model was and basically comparing the prediction to information that was available on the ground. And that indicator was actually a KPI for the project and enabled us to compare ourselves to other types of indicators, macroeconomic indicators that are used by the financial institutions. I'm going to pass on the technical details, but what's interesting is to look a little bit at the results that we've obtained. So with that multilayered AI model, combining information and then enriching it with heterogeneous data, we built a road network that was 60% richer than OpenStreetMap with information on the pavement and the condition of the road that was 80% accurate based on the ground truthing that we've made. And so just to give you a perspective on what it means for a country like Mexico or Peru, usually the Department of Transportation only gets information on 20% of their network. And so we offer them a method at scale that was giving them very granular information on 100% of the roads that they have in their country and basically opening their eyes on most of the investment initiatives that they have to do to connect to the most rural populations. And I'm going to close on the benefits and next steps we see for that type of large-scale initiative, mostly relying on freely accessible data. So if I take Peru, again, we mapped more than 500,000 kilometers of roads, mostly in rural areas where usually you don't really get any information. And this is highly critical, again, to connect to those rural population. Understand if you can plan to put a hospital, schools, or any infrastructure that would actually make the life of these communities better. And that approach is actually, like, six times less expensive than the traditional methods that are used and actually takes only a month to be deployed at a country level, while if you have to do that by ground means, driving a vehicle and take notes on every road of the network, it would take more than 20 years. And what you can see on the left is kind of like an overview of the road quality that we have computed for the country of Peru. And what's interesting about the type of approach for the digital city, but also digital infrastructure initiative is once you have built a digital twin of your network, then you can start analyzing it even further by deploying models that simulate, in that case, that you see on the right, the pressure of the network to climate events. And this is something we've actually done in Tunisia. We've modeled with the same method, a digital twin of the primary and secondary road network, and then built a model to simulate the impact of climate change at the country level and understand which sections of the network would be more or less resilient to those new type of events. And as such, help the government plan for renovation programs that would encompass that notion of climate change in their decision process.