An open-source, flexible, high-performance AI training framework offering multi-agent, multi-method, human-in-the-loop, distributed training for complex ecosystems.
To successfully integrate evolved AI partners with human needs, AI development cannot happen in isolation like it did until now. Different AI agents and users will coexist in a complex ecosystem, training and working together. To tackle this new paradigm of AI development, a more robust, flexible, and future-proof framework was needed: this is why we created Cogment.
Develop tech-heterogeneous components working together regardless of the tech stacks used to develop them. It’s fully compatible with current tools, languages and commonly used techniques.
Cogment doesn't enforce any particular approach to agents implementation. It doesn't favour learning over non-learning agents, nor does it learning techniques over others.
Cogment was designed to allow multiple agents and multiple human users (all "actors") to exist, train, and work together within the same environment, interacting with one another and their environment.
Multiple Reinforcement Learning (RL) agents can use any number or type of reward sources, whether they are from human users, environments (real or simulated), or other agents.
Humans in the loop
Interaction with human users is a core feature of Cogment, at any step of a project from bootstrapping to deployment. There is no enforced limit to the complexity of said interaction.
Several deployments (instances) can contribute to the learning of their agents in a parallel way. Training and use can all happen in a distributed way, running one large algorithm or an array of decentralized agents.
Swap actors in and out from one implementation of an agent to another, or from a human user to another, or from a human user to a trained or untrained agent, and vice-versa.
Virtually no difference between dev and prod versions of a Cogment project; any project can be developed in an iterative way, with any part of it being live-developed so iterations can happen as quickly as possible.
They are trusting us
With AIR, we found a partner who quickly understood our vision for man-machine and machine-machine teaming while allowing us to significantly accelerate our roadmap and unlock new challenges.