As KAI pursued AI Pilot development for its next-generation air combat system, it needed an integrated environment to systematically manage AI models and data. In particular, it required a framework that could reliably run the entire process, from data collection and refinement to model training, validation, deployment, and operations, in a security-critical air-gapped network, and connect to the future validation and use of AI models.

The key challenge was to consistently manage the data and models needed for AI Pilot development while securing both development efficiency and operational stability on a common development platform.

With Cheetah, we built an MLOps platform that supports the entire process, from data collection and refinement to LLM training, validation, deployment, and monitoring, even in an air-gapped environment, and laid the groundwork for validation and use through a future AI Agent platform.

KAI can now systematically manage the data and models needed for AI Pilot development and run the ML lifecycle consistently, even in an air-gapped environment. A common development platform and automation also raised development productivity and operational stability, laying the groundwork for faster validation and release of future AI models and more productive AI agent development.