Azure Updates: AI Foundry; AI Agent Service; Llama 4; Storage

Microsoft CVP Asha Sharma explored new capabilities in Azure AI Foundry to build agentic apps. The company introduced AI Foundry in fall, 2024 as a unified platform for developers and enterprise designers to manage models and tools. It is intended as a modular production system to assemble pre-built components and up to 1800 pre-existing models. 

Microsoft has steadily expanded AI Foundry with the introduction of AI Agent Service, and now the agent framework. Built as an extension of the AI Foundry open-source Semantic Kernel, the agent framework is aimed at coordinating code development. So far, customers like KPMG are using it to orchestrate workflows between specialized agents. The agent framework is supplemented with evaluations, tracing, and A/B testing in an observability suite. Similarly, Microsoft introduced AI Red Teaming Agent, in public preview, as a way to probe models for risks and integrating evaluations into the Microsoft Security PyRIT framework.

Sharma also discussed the release of the first models in the Llama 4 herd for AI Foundry and Databricks. The models, which originate with Meta, are meant to seamlessly integrate text and vision tokens in a unified model backbone. AI Foundry is adding Llama 4 Scout and Maverick models as managed compute offerings. 

“To accommodate a range of use cases and developer needs, Llama 4 models come in both smaller and larger options. These models integrate mitigations at every layer of development, from pre-training to post-training. Tunable system-level mitigations shield developers from adversarial users, empowering them to create helpful, safe, and adaptable experiences for their Llama-supported applications,” Sharma wrote.