Supports GPU-based model training on a Kubernetes platform. Provides distributed training execution, data and checkpoint management, and training performance monitoring (via diverse techniques like benchmarking and judgment)—ensuring efficient GPU utilization, multi-tenant isolation, and reliable large-scale training.

Enable creation, validation, and operation of AI agents, including reusable resources like skills, MCP tools, and knowledge bases. Offer flexible environments for managing agents, registries, and connectors, ensuring centralized control over quality, versioning, and security across deployments.

AI Gateway enables secure, unified management of LLMs, AI Agents, MCP servers, and REST APIs through a single gateway. It standardizes model invocation across external LLM providers and private models, while providing security and operational policies such as guardrails, API key management, and routing.