Skip to main content

Mosaic AI Gateway

Manage and govern all GenAI models across the enterprise

Secure access to any AI model

Mosaic AI Gateway protects your data and GenAI deployments through centralized governance across all models in your enterprise. With support for any model endpoint, including Azure OpenAI, Amazon Bedrock on AWS, and Meta Llama 3, AI Gateway allows you to tap into cutting-edge innovations through a single, powerful query interface. With AI Gateway you can easily test and swap new models without requiring changes to your codebase. It captures usage tracking in Unity Catalog, with permissions enforced to prevent sensitive data leaks. Additionally, with real-time spending insights, AI Gateway keeps your finances in check and your operations streamlined.

Mosaic AI Gateway: Usage tracking and observibility, Payload logging, Gaurdrails, A/B testing and traffic policies, Granular rate limiting, Fine-grained permissions and UC Integration. Mosaic AI Model Serving: External models, databricks Hosted, Agent/chain

Unified access to all AI models

Mosaic AI Gateway offers unified access to a wide range of models, including external LLMs like OpenAI and Meta Llama 3, through a single standard query interface. This eliminates the need to maintain separate systems to manage AI traffic. Enterprises can effortlessly switch between foundation and custom models, ensuring uninterrupted pipelines and simplified maintenance.

Advanced security and governance

Adopting GenAI means navigating complex security protocols, data governance standards and access controls. AI Gateway simplifies this by capturing data flowing through model APIs — such as payload logs, usage tracking and guardrail activation — into Unity Catalog for secure storage, sharing and management. With centralized governance and custom guardrails like content filtration and PII detection, it ensures consistent security and compliance across all endpoints, protecting your organization’s most critical information.

Improved model accuracy

Elevate your model quality with our robust observability tools that capture detailed request and response data. Payload logging enables you to debug, fine-tune and enhance models, improving accuracy and reducing latency. This ensures your AI applications perform at their best and deliver optimal results.

Spending oversight

Our observability tools offer detailed usage tracking and spending oversight. With real-time insights into your AI operations, you can monitor expenses, optimize resource allocation and ensure efficient performance. These tools help you maintain financial control while maximizing the effectiveness of your AI deployments.

Ready to become a data + AI company?

Take the first steps in your data transformation