Datadog's AI Research Lab in Paris hosts PhD candidates Viktoriya Zhukova and Salahidine Lemaachi through France's CIFRE program. Viktoriya researches multimodal timeseries forecasting, leveraging Datadog's vast observability datasets to improve prediction accuracy. Salahidine focuses on world models at the intersection of foundation models and AI observability, including using topological data for anomaly detection. Both contributed to Toto, Datadog's open-source timeseries foundation model presented at NeurIPS 2025 with over 9 million downloads, which has been deployed in production for cloud cost forecasting. The post highlights how Paris's strong academic ecosystem and France's R&D incentives make it an ideal environment for applied AI research.
Table of contents
Improving timeseries forecasting with multimodal dataBuilding world models for AI observabilityWorking in Paris: A research home in a city that inspiresResearch that reaches the real worldSort: