THRSP1065 Theater Tue 17 Nov 14:00 PST
Local when it matters. Cloud when it counts.
Ajay Mungara
Summary
From the recording / Microsoft AI summary
The session made the case for routing agent work between local open-weight models and frontier models. The speaker claimed about 40% savings with under 1% accuracy loss using escalation routing on Intel's own workloads, and said the talk cost about USD 675 to produce with AI agents. He mentioned Intel and Microsoft work on Foundry Local, confidential VMs and Intel TDX, and noted local models commonly expose OpenAI-compatible APIs through vLLM, Ollama and SGLang.
Session description
Microsoft's catalogue description
Learn how agentic AI can combine local and cloud capabilities without choosing one over the other. See how Intel and Microsoft are enabling AI experiences that bring together local models for speed, privacy, and efficiency with Azure intelligence for sophisticated reasoning and enterprise scale.