Wednesday
Room 5
16:20 - 17:20
(UTC+01)
Talk (60 min)
Leveraging Small Language Models for Smarter Workflows
In this session, we’ll explore how to teach function calling to Small Language Models (SLMs). This enables them to automate and execute tasks, and serve as specialized participants in AI workflows. Moreover, for those specific, predefined tasks, SLMs can reach functional parity with much larger language models.
We’ll dive into practical techniques related to using models such as Microsoft’s Phi family, demonstrating how these smaller models can reduce computational overhead while maximizing privacy and performance. Finally, we will explore both running SLMs in Azure, as well as directly embedding them within mobile and client-side applications.
Filip W.
Filip is a founder and maintainer of several popular open source projects, author and a Microsoft Quantum MVP. He specializes in language tools, the Roslyn compiler and quantum computing. Filip is based in Zürich where he works in a medical industry. You can follow him on Twitter @filip_woj.