Datadog, the cloud monitoring and security company listed on Nasdaq, said on June 30, 2026 that it had acquired Adaptive ML, a startup with teams in Paris and New York that helps companies fine-tune their own AI models with reinforcement learning.
Terms were not disclosed. French tech outlet FrenchWeb estimated that a valuation of $100 million to $200 million was likely, based on recent deal multiples, but stressed that the figure was its own estimate rather than a reported price.
Adaptive ML's team will join Datadog AI Research, the company's internal lab. Datadog said the group will work on world models and on post-training large language models for agent tasks in observability, and that it will share updates later on bringing Adaptive ML's capabilities into the Datadog platform.
What Adaptive ML built
Adaptive ML calls its product a reinforcement learning operations, or RLOps, platform. Its software, Adaptive Engine, lets a company take an open model and fine-tune it with reinforcement learning and synthetic data. The models can then be tested against business goals using custom AI judges and A/B tests, and signals from production use are fed back into training so the model improves after each deployment.
The pitch is that enterprises can own and run specialised agents and models rather than relying only on general-purpose models from a provider. When it raised its first round in 2024, coverage of the startup described its aim as helping companies improve their generative AI models based on how users interact with them.
“We started Adaptive to give every enterprise the ability to perpetually improve its own AI. The missing piece was never the algorithm, the hardest part was production scale.”
Why Datadog is buying
In a blog post on the deal, Datadog co-founder Alexis Lê-Quôc argued that off-the-shelf models are easy to deploy but rarely solve complex, domain-specific problems in production, and that lasting value comes from tuning models against an organisation's own live signals.
Datadog holds large volumes of exactly that kind of data from customers' infrastructure, applications and security systems. The two companies plan to build agents and models trained on those observability and security signals. Lê-Quôc described the goal as moving from agents that investigate and send alerts to systems that learn and act.
Datadog's chief scientist, Ameet Talwalkar, said bringing Adaptive ML on board fits the lab's work on turning the company's data into first-party intelligence. Datadog said it spends more than $1 billion a year on research and development. Its AI products include Bits Investigation, Bits Code and Bits Security Analyst, along with the Toto 2.0 model, and the company said its agents have already run hundreds of thousands of investigations for customers.
The founders and the funding
Adaptive ML is led by co-founder and chief executive Julien Launay. In March 2024 it raised a $20 million seed round led by Index Ventures, with ICONIQ Capital, Motier Ventures, Databricks Ventures, IRIS and HuggingFund by Factorial taking part, along with angel investors. At the time it said it was growing its teams in Paris and New York.
FrenchWeb reported that Datadog chief executive Olivier Pomel had invested in Adaptive ML as a business angel. It also said most of the startup's business was in North America and that it had no major French customers at the time of the sale. The companies did not say how many people work at Adaptive ML.
What comes next
The purchase gives Datadog a team with experience in post-training, the stage where a general model is adapted to a narrow task. Launay said the hard part of Adaptive ML's work had never been the algorithm but running it at production scale, and that Datadog's stream of real-world data would let the team build on that foundation.
Datadog presented the deal as part of its research push rather than a new product line. It did not give a timeline for when Adaptive ML's technology will appear in its products.
