Initially, Tardigrade AI developed and hosted its platform on a university supercomputer at Arts et Métiers (ENSAM). However, as the solution developed and the first clients arrived, the need for a more flexible infrastructure became apparent. "The increasing load was becoming difficult to absorb, and we had to manage the infrastructure ourselves.

Together with Valtech, Tardigrade AI rebuilt a cloud-native architecture. The modernization relies on Cloud Run to execute workloads, BigQuery to centralize and exploit data, Cloud SQL and Cloud Storage to manage user accounts and reports, and Cloud Workflows and Pub/Sub to orchestrate the entire process.

AI then transforms this data capital into strategic information. The climate and econometric models used in Adapt, entirely developed in-house, constitute the scientific heart of the platform. However, the results they produce can sometimes be difficult for non-specialists to grasp. To facilitate understanding and offer clear reports to its clients, Tardigrade AI relies on Gemini models: they interpret the results and render them in accessible language. Thanks to this generative component, client reports are automatically produced and enriched with natural language summaries, peril-by-peril insights, and immediate operational adaptation recommendations. "We use generative AI to go from data to insight, the goal being to offer our clients intelligible analyses and directly actionable indicators, thereby facilitating informed decision-making," notes David Kujas.

Empowered by a cloud-native infrastructure, Tardigrade AI continues to enrich its platform. The migration of dashboards to Looker will soon offer dynamic visualizations to finely explore Climate Value at Risk or the cost of inaction. Integration with Google Earth is also expected to foster a better perception of risk. "One meter of water is complicated to grasp. But if I place that meter of water next to a vehicle parked in front of your office in a Google Earth visualization, immediately, you understand better. It allows you to see where the water risks entering and to adapt protection measures," illustrates David Kujas. "This visual realism will help decision-makers objectify the return on investment of preventive measures, such as installing flood barriers or elevating equipment."

This dynamic of innovation extends to R&D: Tardigrade AI is closely monitoring work by Google Labs on modeling complex climate phenomena (storms, hail) to integrate them as soon as they become available. For David Kujas, these advances could further strengthen analysis precision. "Indeed, the adoption of Google Cloud clearly marked a turning point in our evolution: it allowed us to accelerate development, expand our international presence, and open new perspectives by combining Google Cloud's innovation strength with our own." A view shared by Valtech, for whom this project responds to "a major social and environmental necessity," recalls Serge Biscard, Managing Director of Valtech France. And it is precisely this course that David Kujas intends to maintain: "Google Cloud allows us to move faster, further, and to focus our efforts on what counts: helping organizations understand their climate exposure and build their financial resilience."