Remanence
Category: AI Infrastructure
Paris-based continual-learning platform that turns enterprises' AI-agent execution traces into RL training environments and post-trains open-weight models the customer owns. Remanence was founded in 2026. The company is led by Pierre-Louis Cedoz. Based in Paris, France. Team size: 2-10. Total funding raised: $30M. Latest round: Seed. Key investors include EQT Ventures & Growth, FirstMark Capital, Zero Prime Ventures, Emblem, Drysdale Ventures, Global Founders Capital, Yellow, Kima Ventures, OPRTRS CLUB, Darkmode Ventures, Cleo Ventures, Olivier Pomel (Datadog).
- Founded
- 2026
- Headquarters
- Paris, France
- Team size
- 2-10
- Total funding
- $30M
Value proposition
Companies stop renting intelligence from closed model providers and instead own specialised open-weight models that keep improving as their agents generate more execution data, with better accuracy, lower latency and lower cost.
Products and solutions
Continual-learning platform that connects to a company's existing agents, converts execution traces into tailored training environments, verifiers and tasks, post-trains open-weight models with reinforcement learning, and serves them via managed inference, in the customer's cloud or VPC, or on-prem.
Unique value
Training environments are generated automatically from the customer's own production agent traces rather than from a generic set. The customer owns the model weights and the data can stay in its own infrastructure.
Target customer
Enterprises and AI-native companies running AI agents in production, including regulated industries that need VPC or on-prem deployment.
Industries served
Legal, customer support, enterprise software (design partners in Europe and the US)
Technology advantage
Founders led long-horizon RL post-training and environment generation for H Company's open-weight Holo computer-use models, which the company says outperformed leading open and closed models on computer-use benchmarks. Team backgrounds include Google DeepMind, Meta and InstaDeep. Funding includes a dedicated GPU cluster.
How they differentiate
Synthesizes training environments from the execution traces of agents already in production and ties long-horizon RL post-training and inference into one loop. The customer owns the weights, with flexible deployment (own cloud, VPC, on-prem).
Main competitors
Applied Compute, Bespoke Labs, Mercor
Key partnerships
Design partners in legal, customer support and enterprise software across Europe and the US (names undisclosed).
Major milestones
Company created in Oct 2026 by three H Company alumni, emerged from stealth on 2026-11-17 with a $30M seed co-led by EQT Ventures & Growth and FirstMark.
Growth metrics
Came out of stealth on 17 Nov 2026 with a $30M seed. Team of roughly 2-10 at announcement, hiring researchers and engineers in Paris, London and San Francisco.
Market positioning
Early-stage European enterprise AI infrastructure player in the continual-learning and post-training space. Bets on a hybrid enterprise stack of rented frontier models plus owned, specialised open-weight models.
Geographic focus
Europe and US; hiring mainly in Paris, with early presence in London and San Francisco.
About Pierre-Louis Cedoz
Ex-H Company CTO (from early 2026; research lead since 2024; left in early Sept 2026), led post-training of the open-weight Holo models; Google DeepMind background; Stanford University.
Latest news about Remanence
More AI Infrastructure companies
Official website: https://www.remanencelabs.ai