Adaptive ML
Adaptive ML built Adaptive Engine, an enterprise platform that fine-tunes open-source LLMs with reinforcement learning for Fortune 500 customers in financial services, telecoms, and insurance; acquired by Datadog in July 2026.
- Company typePrivate
- Founded2023
- HeadquartersParis, France
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Adaptive ML does
Adaptive ML is a Paris-headquartered enterprise AI company founded in September 2023 by Julien Launay, formerly of Hugging Face, alongside other contributors to the open-source Falcon LLM family. The company develops Adaptive Engine, an enterprise Reinforcement Learning Operations (RLOps) platform that enables large organizations to fine-tune open-source large language models (Llama, Mistral, Gemma) using RL methods including PPO, DPO, RLAIF, RLEF, GRPO, and GSPO. The platform is composed of three modules — Adapt (training via reinforcement learning, synthetic data generation, and AI judges), Evaluate (A/B testing and bespoke AI judges against business KPIs such as RAG faithfulness, escalation rate, and policy adherence), and Serve (deployment with Pareto frontier exploration for model right-sizing). The underlying proprietary codebase, Adaptive Harmony, is written in Python for high-level logic and Rust for distributed coordination, and is exposed through a developer SDK and Python documentation.
The company serves regulated, large-enterprise customers across financial services (Manulife for underwriting and sales advisory; an unnamed Fortune 100 financial firm for RAG on 10-K reports), telecommunications (AT&T across 50+ use cases including text-to-SQL, customer support, and fraud detection; SK Telecom for multilingual content moderation), and healthcare (CCS for patient support function-calling), with Deloitte as an implementation partner and HPE Private Cloud AI plus NVIDIA DGX B200 certification providing OEM distribution. Adaptive ML raised a $20 million seed round in March 2024 led by Index Ventures at a reported $100 million valuation, and appointed a dedicated CRO and CMO in April 2026 to scale its enterprise field sales motion. On July 1, 2026, Adaptive ML was acquired by Datadog, with the team folded into Datadog AI Research to advance work on world models and agentic LLM post-training for observability.
Adaptive ML's revenue model is enterprise subscription-based, typically structured as multi-year agreements sold through direct enterprise field sales and OEM partners, with pricing not publicly disclosed. Quantified customer outcomes include a 100% accuracy improvement (27% to 58% win rate vs GPT-4o) on financial RAG with Llama 3.1 8B, approximately 80% annual cost reduction migrating a Fortune 500 customer operations workload from a frontier model to an 8B specialist, and a 51% win rate vs a leading closed-source LLM in telco document RAG at AT&T.
Adaptive ML firmographics
Firmographics- Name
- Adaptive ML
- Legal name
- Adaptive ML, Inc.
- Website
- https://adaptiveml.co
- Company type
- Private
- Founded year
- 2023
- Operating status
- Acquired
- Headcount range
- 11–50 employees
- Short description
- Adaptive ML built Adaptive Engine, an enterprise platform that fine-tunes open-source LLMs with reinforcement learning for Fortune 500 customers in financial services, telecoms, and insurance; acquired by Datadog in July 2026.
- Ownership category
- akta.pro rank
Adaptive ML industry classification
Industry- Product category
- Enterprise LLM Fine-Tuning Platform
- NAICS
- Software Publishers (5132), Custom Computer Programming Services (541511)
- SIC
- Services-Computer Programming, Data Processing, Etc. (7370), Services-Prepackaged Software (7372)
- akta.pro primary industry
- RLHF, Human Feedback & Evaluation Data Tools (HDAAALAJ)
Keywords
Where Adaptive ML is headquartered
LocationHeadquarters
- HQ city
- Paris
- HQ country
- France
- HQ region
- Europe
Offices3 records
Markets served
Adaptive ML business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Marketing or Sales, Infrastructure, Operations
Revenue model
- Adaptive Engine Platform Subscription: Subscription-based access to Adaptive Engine platform for model fine-tuning, evaluation, and deployment. Typically multi-year enterprise agreements. Sold directly to enterprises and via OEM partners (HPE PCAI).
Go-to-market motion4 records
Distribution channels4 records
Marketing channels7 records
Adaptive ML product offering
Product offeringCore offering
Adaptive ML provides an enterprise software platform, Adaptive Engine, that enables companies to fine-tune open-source large language models using reinforcement learning methods (RLHF, RLAIF, RLEF, PPO, DPO, GRPO). The platform combines a training module (Adapt), an evaluation module (Evaluate), and a deployment module (Serve) with custom AI judges and synthetic data generation to allow enterprises to continuously improve domain-specific AI models from production feedback rather than relying on third-party LLM APIs.
Product overview
Adaptive ML offers Adaptive Engine, an enterprise reinforcement learning operations (RLOps) platform that enables companies to build, own, and continuously improve specialized AI models. The platform consists of three integrated modules: Adapt (RL fine-tuning using PPO, DPO, GRPO methods with synthetic data and AI judges), Evaluate (A/B testing and bespoke AI judges), and Serve (model deployment with rightsizing). Supporting the platform is Adaptive Harmony, a Python-based codebase for preference tuning built with Rust for distributed operations. The company focuses on enterprise use cases including RAG, text-to-SQL, customer support, and AI agents, serving clients like AT&T, Manulife, HPE, and SK Telecom. Following acquisition by Datadog in July 2026, Adaptive ML operates within Datadog AI Research to advance world models and agentic LLM post-training for observability.
Differentiator
Problem solved
Functional benefit
Brands
- Adaptive Engine: Enterprise platform enabling companies to continuously improve their large language models by learning from user interactions using reinforcement learning techniques.
- Adaptive Harmony
Products and services
- Adaptive Engine Enterprise Reinforcement Learning Operations (RLOps) platform enabling companies to fine-tune, evaluate, and deploy open-source large language models using RLHF/RLAIF/RLEF methods. Built around three integrated modules — Adapt (training), Evaluate (AI judges and A/B benchmarking), and Serve (deployment and Pareto-frontier model rightsizing) — and designed for continuous model improvement from production feedback, custom AI judges, and synthetic data generation.
- Adaptive Harmony Proprietary Python (with Rust coordination layer) codebase built from the ground up for preference tuning. Implements PPO, DPO, RLOO, and Constitutional AI methods in a few lines of high-level Python while using Rust for distributed coordination, delivering production-grade reliability and performance for novel RL research.
Quantifiable outcome
- 100% accuracy improvement in financial RAG (Llama 3.1 8B base model win rate vs GPT-4o improved from 27% to 58%)
- +6 more outcomes
Companies that use Adaptive ML
Customer profileNamed customers5 records
Segments5 records
Ideal customer profiles4 records
Adaptive ML technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration4 records
AI capability10 records
Feature8 records
Adaptive ML partnerships and signals
Strategic signalPartnerships
Seven partnerships are on record, tiered core and flagship.
- DeloittecoreDeloitte partnered with Adaptive ML and CCS in 2025 to develop AI systems for healthcare patient support. Deloitte is deploying Adaptive ML's reinforcement fine-tuning capabilities as part of a broader enterprise AI transformation engagement for CCS, a chronic care management organization.
- ManulifeflagshipMulti-year strategic partnership making Adaptive Engine Manulife's reinforcement learning operations layer. Manulife is deploying Adaptive ML's RL fine-tuning technology across its global enterprise AI platform for automating underwriting quotes, complex process execution, and sales professional advisory. Manulife is the #1 ranked life insurer for AI maturity in the Evident AI Index and has deployed 140+ AI use cases as of 2025.
- SK TelecomcoreSK Telecom collaborated with Adaptive ML to fine-tune Gemma 3 4B and other open models for multilingual customer support content moderation in Korean and English. The tuned models achieved performance equal to or better than GPT-4.1, o4-mini, GPT-4o, and Claude 3.7 Sonnet at a fraction of the size and latency, enabling low-latency, on-premises content moderation.
- NVIDIAcoreAdaptive Engine was officially certified for NVIDIA GBX B200 (Blackwell architecture DGX systems), joining NVIDIA's ecosystem of enterprise AI partners as an NVIDIA-Certified System. The certification was announced at NVIDIA GTC Paris, enabling Adaptive Engine to run on NVIDIA DGX B200 systems delivering 3x training performance and 15x inference performance vs DGX H100.
- AT&TflagshipAT&T engaged Adaptive ML deploying Adaptive Engine as their reinforcement tuning platform for open-source models. AT&T is deploying fine-tuned reasoning models across 50+ use cases including AskData (text-to-SQL), customer support, call summarization, and document RAG. AT&T also expanded AI collaborations with Adaptive ML for fraud detection as part of a broader enterprise AI services diversification strategy.
- HPE (Hewlett Packard Enterprise)coreHPE partnered with Adaptive ML to offer Adaptive Engine within HPE's Private Cloud AI (PCAI) offering, co-developed with NVIDIA. PCAI provides enterprises with private cloud GenAI infrastructure and Adaptive Engine is included as part of the validated tooling ecosystem, enabling enterprises to accelerate GenAI projects from pilot to production while keeping data private and secure.
- CCS (Chronic Care Management)coreCCS, a US chronic care management organization providing medical supplies and services to diabetes patients, engaged Adaptive ML and Deloitte in 2025 to develop AI for patient support. The collaboration fine-tuned Llama 3.2 3B via Adaptive Engine for function-calling AI agents, achieving proprietary model-level accuracy at 90%+ lower latency on a single H100 instance.
Scale indicators10 records
Recent moves6 records
Expansion highlights6 records
Adaptive ML competitors and assessment
Company assessmentBroad incumbents
- Hugging Face: Hugging Face is the dominant open-source model and dataset hub and offers training/fine-tuning tooling (TRL, AutoTrain). Adaptive ML's CEO Julien Launay previously built fine-tuning capabilities there; both companies serve enterprise LLM builders, though Hugging Face's offering is broader (models, data, spaces) while Adaptive ML is specialized in RL post-training.
- MosaicML (now part of Databricks): MosaicML provided an enterprise platform for training and fine-tuning large models on proprietary data before being acquired by Databricks. It is the closest large-scale analog to Adaptive Engine in purpose-built enterprise model training/fine-tuning, and the fact that both MosaicML (Databricks) and Adaptive ML (now Datadog) ended up inside data/AI platform incumbents underscores the category convergence.
Direct peers
- Weights & Biases: Weights & Biases is an MLOps/LLMOps platform providing experiment tracking, model evaluation, sweeps and reinforcement-learning tooling used by AI teams to fine-tune and evaluate LLMs. It overlaps directly with Adaptive Engine's Evaluate module and the broader RLOps workflow (training, benchmarking, deployment).
- Scale AI: Scale AI is a leading data-labeling and RLHF services provider whose Donovan/Spellbook platforms target enterprise model fine-tuning and evaluation. It competes most directly with Adaptive ML's RLAIF/custom AI judges capability by offering human-in-the-loop and AI-feedback data generation for enterprise LLM training.
- Together AI: Together AI offers a cloud platform for fine-tuning and serving open-source LLMs, with dedicated fine-tuning APIs and inference infrastructure targeting enterprise developers. It competes with Adaptive ML on the same open-model enterprise fine-tuning proposition and shares the same base-model-agnostic positioning.
- Lamini: Lamini is an enterprise-focused platform for fine-tuning and serving proprietary LLMs on company data, with emphasis on accuracy, hallucination reduction and private deployment. It targets the same 'build your own specialist model' buyer as Adaptive Engine, particularly in regulated industries.
- Snorkel AI: Snorkel AI provides a data-centric AI platform for programmatic labeling, weak supervision and RLHF-style feedback on enterprise data. It competes in the same enterprise fine-tuning-tooling niche as Adaptive ML, with comparable enterprise buyers in financial services and regulated verticals.
- Labelbox: Labelbox provides a data-centric AI platform covering labeling, evaluation and RLHF workflows for enterprise model training. It overlaps with Adaptive ML's RLAIF and AI-judge capabilities for building enterprise preference datasets and evaluation pipelines.
Emerging players
- Anyscale: Anyscale (built on Ray) provides distributed compute infrastructure used by AI teams to run large-scale training, fine-tuning and RL workloads. It is an enabling layer for the same RL post-training workflows that Adaptive ML's Adaptive Engine orchestrates, and is often evaluated alongside RL/HF toolchains.
- Argilla: Argilla is an open-source data labeling and feedback platform widely used for RLHF/RLAIF datasets, supporting human and AI feedback workflows. It targets a similar developer persona building preference data pipelines for fine-tuning LLMs, and serves as a lighter-weight alternative to Adaptive Engine's Adapt module.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat6 records
Key risks6 records
Key highlights7 records
Customer concentration
Adaptive ML social profiles
Digital presenceAdaptive ML financial estimates
Financial estimateRevenue estimate
Valuation estimate
Adaptive ML leadership team
Management profileNumber of profiles
Profiles8 records
Adaptive ML funding detail
Funding detailFunding overview
Funding rounds2 records
Investors8 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Adaptive ML M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Adaptive ML
What does Adaptive ML do?
Adaptive ML provides an enterprise software platform, Adaptive Engine, that enables companies to fine-tune open-source large language models using reinforcement learning methods (RLHF, RLAIF, RLEF, PPO, DPO, GRPO). The platform combines a training module (Adapt), an evaluation module (Evaluate), and a deployment module (Serve) with custom AI judges and synthetic data generation to allow enterprises to continuously improve domain-specific AI models from production feedback rather than relying on third-party LLM APIs.
Is Adaptive ML a public or private company?
Adaptive ML is a private company. It is classified as corporate owned and is currently acquired.
When was Adaptive ML founded?
Adaptive ML was founded in 2023. It employs 11 to 50 people.
Where is Adaptive ML based?
Adaptive ML is headquartered in Paris, France, in the Europe region.
How does Adaptive ML make money?
One revenue line is on record: adaptive Engine Platform Subscription.
Who are Adaptive ML's main competitors?
Broad incumbents on record are Hugging Face and MosaicML (now part of Databricks). Direct peers are Weights & Biases, Scale AI, Together AI, Lamini, Snorkel AI and Labelbox. Emerging players are Anyscale and Argilla.
Does Adaptive ML have an API?
Yes. Adaptive ML provides documentation for developers to integrate and build with Adaptive Engine, including guidance on custom reward functions, training methods (PPO, DPO, GRPO), and evaluation frameworks. The documentation covers Reinforcement Learning from Execution Feedback (RLEF) through custom reward servers. Developer documentation is at docs.adaptive-ml.com.
What industry is Adaptive ML in?
Adaptive ML's product category is Enterprise LLM Fine-Tuning Platform. Its primary akta.pro industry code is HDAAALAJ, RLHF, Human Feedback & Evaluation Data Tools. Its NAICS code is 5132 and its SIC code is 7370.