AgileRL Raises $7.5M for AI Training Platform

London-based startup AgileRL has raised $7.5 million in seed funding to expand its AI training software platform. The company was co-founded in 2023 by Param Kumar and Nicholas Ustaran-Anderegg.

AgileRL focuses on reinforcement learning (RL), an AI training technique where systems learn by trying actions and improving based on feedback. While this technology dates back to the 1950s, it is currently experiencing a renaissance in AI labs.

The startup has created a platform called Arena. This software allows engineers and data scientists to drop their AI models into simulations, fine-tune them before deployment, and monitor them while running.

According to the company, the platform has been downloaded more than 300,000 times. It is currently used by major corporations including Airbus, IBM, and JPMorgan.

The funding round was led by Fusion Fund, with participation from Flying Fish, Octopus Ventures, Entrepreneur First, and Counterview Capital. AgileRL plans to use the capital to open an office in San Francisco and hire more than a dozen people in engineering and go-to-market roles.

AgileRL has secured $7.5 million in seed funding to accelerate the development of its proprietary software. The startup, based in London, aims to simplify the complex process of AI training for businesses.

The core of their offering is a platform named Arena. It is designed to be a comprehensive environment where developers can manage the entire lifecycle of an AI model. Specifically, Arena allows users to:

• Drop AI models into the platform
• Run simulations
• Fine-tune models prior to deployment
• Monitor models while they are actively running

The company argues that their solution speeds up AI development because the training tools are consolidated in one place and are "off the shelf." This contrasts with the alternative of setting up a proprietary AI lab from scratch, which is often resource-intensive.

Regarding business model, AgileRL offers a free tier that provides users with access to a limited amount of training credits. For larger needs, paid tiers are available for businesses and professionals, alongside custom licenses for enterprises.

AgileRL is betting on reinforcement learning (RL), a technique that has roots dating back to the 1950s. Unlike other AI methodologies, RL involves systems learning one step at a time by trying actions and improving based on the feedback they receive.

Param Kumar, the CEO of AgileRL, provided insight into the current market trends regarding AI training technologies. He noted a significant shift following the launch of ChatGPT in late 2022.

"After ChatGPT launched in late 2022, companies moved their budgets from working on RL to focus on transformers," Kumar stated. Transformers are the technology that underpins large language models, learning patterns from large datasets all at once.

However, Kumar believes the industry is realizing the limitations of this approach. "Now, he says more companies are realizing that transformers can only get them so far." He explained that while transformers are great, they are essentially large statistical models.

Kumar elaborated on the necessity of combining technologies: "The reality is you will need to layer on RL on top of that, because there's only so much you can infer from the data."

To illustrate the practical application of reinforcement learning, Param Kumar provided a specific example involving robotics. He described a scenario where a robotic arm is tasked with moving a ball from one table to another.

Kumar explained that this movement is not a single action but can be broken down into many smaller tasks. These include grasping the ball, lifting the arm, and moving the joint. He noted that AgileRL's platform allows engineers to set parameters to improve at these specific tasks.

The startup has seen significant adoption since its inception. The platform has been downloaded more than 300,000 times. It has also attracted high-profile enterprise clients, with usage confirmed by companies including Airbus, IBM, and JPMorgan.

With the $7.5 million seed round led by Fusion Fund, AgileRL is positioned for rapid expansion. The capital will support the opening of a new office in San Francisco. Furthermore, the startup plans to hire more than a dozen people to fill engineering and go-to-market roles.

AgileRL represents a growing trend in the AI sector: the return to reinforcement learning to solve complex problems that large language models cannot handle alone. By providing an accessible, off-the-shelf platform like Arena, the startup is lowering the barrier to entry for companies looking to implement advanced AI training.

With strong backing from investors like Fusion Fund and a user base that includes some of the world's largest corporations, AgileRL is well-positioned to capitalize on the renewed interest in RL. The planned expansion into San Francisco and increased hiring signal a commitment to scaling their operations globally.

As the AI landscape continues to evolve, the ability to combine the statistical power of transformers with the step-by-step learning of reinforcement learning will likely define the next generation of AI applications. AgileRL aims to be at the forefront of this technological convergence.