Ameet Talwalkar, Datadog's Chief Scientist, emphasized the importance of these projects in developing research ideas and collaborating with engineering teams to benefit customers.

Toto is an open-source time series foundation model trained on Datadog's internal telemetry metrics, designed for observability with zero-shot forecasting capabilities.

Toto outperforms existing time series foundation models, enabling instant anomaly detection and capacity planning across billions of ephemeral time series without the need for per-series tuning.

This benchmark provides a standardized testing ground for researchers, capturing characteristics like scale, sparsity, and cold-start issues present in production telemetry data.

The initiative reflects Datadog's commitment to tackling real-world challenges while contributing to the broader research community.

To enhance its AI monitoring capabilities, Datadog recently acquired Metaplane, an AI observability start-up, aiming to provide reliable solutions for customers.

BOOM, the largest public benchmark for observability metrics, features 350 million observations across 2,807 multivariate series, addressing unique challenges such as scale and sparsity in production telemetry.

Datadog AI Research aims to bridge academic developments with practical implementations in cloud observability and security, promising regular contributions to the research ecosystem.

Both Toto and BOOM are available for immediate download under a permissive open-source license, encouraging community engagement for further research.

On May 21, 2025, Datadog launched two significant AI research initiatives, Toto and BOOM, through its Datadog AI Research division.

Datadog AI Research is dedicated to continuously contributing to the research community by publishing findings and open-sourcing model artifacts.

This AI research initiative aligns with Datadog's expansion into government markets, as the company has achieved FedRAMP High authorization to meet stringent security requirements for federal agencies.