Metis

Metis operates by utilizing first-party data to create a continuously learning agent stack that improves over time. It generates training signals and employs both offline and online reinforcement learning, ensuring agents can complete complex workflows reliably.

Metis is powered by a proprietary intelligence framework designed for enterprise agents. This framework includes a post-training and continual-learning layer that enhances the capabilities of agents operating in real-world environments.

Metis's main features include continuous learning from first-party data, real-time evaluation of agents against actual tasks, and the ability to preserve context while completing multi-step workflows. This ensures higher reliability and throughput for enterprise applications.

Metis employs reinforcement learning techniques, both offline and online, to enhance agent performance. This approach allows agents to learn from interactions and improve their decision-making capabilities over time.