Adaptive ML

Adaptive ML develops infrastructure for enterprises to specialize language models through reinforcement learning. Its Adaptive Engine platform combines synthetic training data, model tuning, bespoke AI judges, A/B testing and production feedback. Enterprises can deploy it on premises or in secure cloud environments for applications including enterprise search, natural-language database queries and customer support. AT&T uses the platform for specialized telecom workflows.

Founded in 2023 by LightOn and Hugging Face alumni, its founding team comprised Julien Launay, CEO at acquisition, Baptiste Pannier, Daniel Hesslow, Axel Marmet, Olivier Cruchant and Alessandro Cappelli. Its Harmony compute backend unifies inference and reinforcement-learning training: the same model weights support both rollout generation and training, avoiding the synchronization burden of separate model stacks. Datadog completed its acquisition of Adaptive ML in 2026, bringing the team into Datadog AI Research to advance world models and agentic language-model post-training for observability.

In June 2026, the company reported that Adaptive Engine had processed trillions of tokens across enterprise deployments during the preceding year. Its May 2026 AT&T case study described call summarization at approximately 900,000 transcripts per day. Before the acquisition, Adaptive ML raised a $20 million seed round led by Index Ventures in 2024.