Where is Adaptive ML headquartered?

Adaptive ML is headquartered in New York, United States; Paris, France. The company has a team of 51-100 people.

What does Adaptive ML do?

Adaptive ML operates in the AI Infrastructure sector. Adaptive ML builds a Reinforcement Learning Operations (RLOps) platform enabling enterprises to build, own, and deploy specialized AI models using reinforcement learning and preference tuning on open-source LLMs. Enables enterprises to build, own, and continuously improve specialized AI models using open-source LLMs tuned with reinforcement learning, outperforming proprietary frontier APIs while retaining full data ownership and model control.

When was Adaptive ML founded?

Adaptive ML was founded in 2023. It is based in New York, United States; Paris, France.

Who is the CEO of Adaptive ML?

Julien Launay is the CEO of Adaptive ML. The company was founded in 2023.

Where is Adaptive ML located?

Adaptive ML is headquartered in New York, United States; Paris, France.

What funding stage is Adaptive ML at?

Adaptive ML is currently at the Acquired stage.

How many employees does Adaptive ML have?

Adaptive ML has approximately 51-100 employees.

Adaptive ML

Category: AI Infrastructure

Adaptive ML builds a Reinforcement Learning Operations (RLOps) platform enabling enterprises to build, own, and deploy specialized AI models using reinforcement learning and preference tuning on open-source LLMs. Adaptive ML was founded in 2023. The company is led by Julien Launay. Based in New York, United States; Paris, France. Team size: 51-100. Total funding raised: $20M. Latest round: Seed. Key investors include Index Ventures, ICONIQ Capital, Motier Ventures, Databricks Ventures, IRIS, HuggingFund by Factorial.

Founded: 2023
Headquarters: New York, United States; Paris, France
Team size: 51-100
Total funding: $20M

Value proposition

Enables enterprises to build, own, and continuously improve specialized AI models using open-source LLMs tuned with reinforcement learning, outperforming proprietary frontier APIs while retaining full data ownership and model control.

Products and solutions

Adaptive Engine (RLOps platform): ADAPT (Bootstrap with Reinforcement Learning — synthetic data generation, fine-tuning with RL), EVALUATE (Evaluate with bespoke AI judges — custom evaluation aligned to business outcomes, A/B testing), SERVE (Serve & Adapt — production feedback loop, real-time metric tracking, continuous model improvement). Adaptive Harmony (in-house unified preference tuning stack for inference, training, and RL). Use-case specific solutions: Enterprise Search, Business Intelligence, Customer Support agents.

Unique value

RLOps platform that closes the loop between production feedback and model training — allowing small, open models to exceed frontier API performance through continuous reinforcement learning, with full enterprise ownership and data privacy.

Target customer

Large enterprises and Fortune 500 companies deploying LLMs for customer support, enterprise search, business intelligence, text-to-SQL, call summarization, document RAG, and content moderation.

Industries served

Telecommunications, Financial Services/Insurance, Customer Support, Enterprise Software, Technology, Cloud Observability & Security (post-acquisition)

Technology advantage

Deep expertise in reinforcement learning from founders who contributed to Falcon and BLOOM open-source models. Adaptive Harmony unified RL infrastructure. Proprietary synthetic data generation for RL training. Certified for NVIDIA GBX B200 Blackwell architecture. NVIDIA NeMo integration for RAG accuracy. Demonstrated 51% win rate vs GPT-4o with fine-tuned Llama 3.1 8B on telco RAG tasks.

How they differentiate

First dedicated Reinforcement Learning Operations (RLOps) platform purpose-built for enterprise LLM deployment. Focuses on post-training (RL finetuning) vs. pre-training. Automated RLAIF (Reinforcement Learning from AI Feedback) replacing expensive human annotation. Adaptive Harmony unified codebase enables preference tuning with just a few lines of code. Proprietary production feedback loop that improves models with every user interaction.

Main competitors

Weights & Biases (MLOps/experiment tracking), Unsloth (fine-tuning optimization), Predibase (low-code fine-tuning platform), Together AI (fine-tuning + inference cloud), IBM Watsonx (enterprise AI governance + tuning)

Key partnerships

NVIDIA (certified system partner for GBX B200 Blackwell architecture, NeMo integration), Google/DeepMind (Gemma case study collaboration), Datadog (acquired June 2026, joining Datadog AI Research)

Notable customers

AT&T (50+ use cases, deployed Adaptive Engine as RL tuning platform), SK Telecom (tuned Gemma 3 4B for multilingual content moderation), Manulife (multi-year agreement for enterprise AI scale), Aïkan (Juribot insurance document chatbot)

Major milestones

Founded 2023 by ex-Hugging Face and LightOn researchers, $20M Seed round led by Index Ventures at ~$100M valuation (March 2024), First version of Adaptive Engine shipped, AT&T selected Adaptive Engine as enterprise RL tuning platform (2024-2025), SK Telecom published Gemma 3 case study (2025), Manulife multi-year agreement (2025), NVIDIA certified system partner (2025), Acquired by Datadog to join Datadog AI Research (June 2026)

Market positioning

Specialized RLOps platform bridging the gap between off-the-shelf LLMs and production-ready enterprise AI. Positioned as the "industrialized intelligence layer" — enabling enterprises to own their AI stack rather than rely on third-party API providers. Acquired by Datadog in June 2026 to power frontier AI infrastructure for cloud observability and security.

Geographic focus

North America (New York HQ) and Europe (Paris office), with global enterprise reach via customers like AT&T, SK Telecom, and Manulife

About Julien Launay

Ex-Hugging Face Extreme-Scale Research Lead; Ex-LightOn Extreme-Scale Research Lead. Background in engineering, physics, and mathematics from École Normale Supérieure Paris-Saclay.

Official website: https://www.adaptive-ml.com