Beyond ChatGPT: Remanence Secures $30M to Make AI More Personalised

Avatar photo

As companies move from experimenting with artificial intelligence to using it for everyday business operations, the cost and control of AI systems are becoming increasingly important. Paris based startup Remanence believes businesses should not have to depend entirely on expensive, general purpose models developed by major technology companies. Its approach is to help organisations transform existing open weight AI models into specialised systems trained around their own workflows, data and operational needs.

The company has raised $30 million in seed funding to develop this vision, marking a significant investment in the growing market for enterprise AI infrastructure.

Major Investors Back the Startup

The funding round was co-led by EQT Ventures and FirstMark, with participation from Zero Prime, Emblem, Drysdale, Global Founders Capital and Yellow.

The round also attracted prominent angel investors, including Datadog CEO Olivier Pomel, Pigment co-founder Romain Niccoli, and Synthesia co-founders Victor Riparbelli and Steffen Tjerrild.

The capital will support research and product development, investment in computing infrastructure, including a dedicated GPU cluster, and recruitment. Remanence plans to build its team primarily in Paris while establishing an early presence in London and San Francisco.

Turning AI Experience Into Specialised Models

Most businesses currently use general purpose AI models to power their digital assistants and automated workflows. Although these systems can perform a wide range of tasks, they may not fully understand a company’s internal processes, software tools or specific operational requirements.

Remanence wants to address this gap by helping organisations customise open weight models. These models provide access to their underlying parameters, allowing businesses to adapt them for particular applications instead of relying exclusively on externally hosted proprietary systems.

The startup’s platform turns real world AI agent activity into learning opportunities. It analyses execution records to identify where an agent succeeds, where it makes mistakes and which tasks require improvement.

The platform then converts these records into interactive training environments that reproduce the company’s workflows. AI models can practise tasks in these simulated environments, receive feedback and improve their performance.

Reinforcement Learning at the Centre

Remanence uses reinforcement learning and advanced post training techniques to improve the models. During training, successful actions are rewarded, while unsuccessful behaviour provides information that can guide further adjustments.

Its official platform describes a four stage process: trace analysis, environment synthesis, long horizon reinforcement learning and inference.

First, the system examines an organisation’s agent execution data to identify weaknesses. It then creates simulated environments and tasks based on those records. Open weight models are trained against the company’s existing performance baseline before being deployed through managed infrastructure or the customer’s own cloud.

This creates a continual learning cycle in which new production experiences can become training material for future improvements.

Giving Businesses Greater Control Over AI

A major part of Remanence’s proposition is ownership. Rather than simply paying for access to an external AI model, companies can retain control over the specialised model weights developed through the process.

The startup says its technology can help businesses improve accuracy and reduce latency and operating costs on their specific tasks. Its website also highlights deployment options within a company’s private cloud or on premises infrastructure, an important consideration for organisations handling sensitive information or operating in regulated sectors.

The platform is designed for businesses that want to improve existing AI agents without relying solely on prompt changes or adjustments to the tools surrounding them.

Experienced Team Targets Enterprise AI

Remanence was founded by Pierre-Louis Cedoz, Hamza Benchekroun and Axel Moyal. The founders previously worked at French AI startup H Company, where they gained experience in developing and post training advanced AI models. Their professional backgrounds also include organisations such as Google DeepMind, Meta and InstaDeep.

CEO Pierre-Louis Cedoz said the founders had experienced both the potential of post training and the difficulty most businesses face in accessing the necessary expertise. The company’s ambition is to help organisations turn their own agent data into specialised models that improve with use.

Remanence is already working with design partners in legal services, customer support and enterprise software across Europe and the United States.

A Shift Towards AI Ownership

The company is entering a market where businesses are increasingly evaluating the cost, performance and ownership implications of their AI infrastructure. While proprietary models from major AI laboratories remain important, specialised open weight systems could offer an additional option for organisations with demanding or repetitive workflows.

Remanence is betting that a company’s operational experience can become a valuable source of competitive advantage when converted into models it controls. Its next challenge will be demonstrating that these customised systems can deliver reliable results, lower costs and measurable improvements in real business environments.

Total
0
Shares
Previous Post

Light Takes the Lead: Qluster Secures €4.7M for Its Quantum Computing Ambitions

Next Post

Beyond Chatbots: Model ML Secures $250M to Build AI Agents for Finance

Related Posts