Beyond the Prompt: Tuning Fabric Data Agents for Flawless Results

Data Engineers in Toronto August 2026 Semimonthly Meeting

Topic: Beyond the Prompt - Tuning Fabric Data Agents for Flawless Results

Abstract:
As organizations race to deploy AI Agents, it quickly becomes clear that standard prompt engineering isn't enough. To achieve production-grade accuracy, you have to go beyond the prompt and fundamentally change how you feed your data layers to the AI.

In Microsoft Fabric, your choice of data artifact directly dictates your agent's reasoning style and native language.

In this session, we will deep-dive into how Fabric Data Agents interact with four distinct data paradigms: Semantic Models, Ontology, Data Warehouses, and Lakehouses.

Moving past high-level theory, we will practically map the three pillars of data agent optimization across the data paradigms:
Agent Instructions (System Prompts)
Data Source Descriptions (Grounding)
Example Queries (Few-Shot Prompting)

You will walk away with an actionable blueprint showing you exactly how to write context rules, embed semantic metadata, and supply technical query examples to ensure your Fabric Data Agents deliver flawless, trusted results every time.

Speaker: Thimantha Vidanagamage, Data Engineer at OMERS, Canada

Speaker Profile:
Thimantha has over 8 years of experience in Data and AI. He currently works as a Senior Data Engineer at OMERS in Canada, where he plays an active role in the organization’s adoption of Microsoft Fabric. He holds multiple Microsoft certifications, including Microsoft Certified: AI Transformation Leader and Fabric Data/Analytics Engineer Associate. Thimantha is also a co-organizer of the Toronto Data Professionals Community.

The meeting is over Microsoft Teams, and the joining link is https://vip.dataengineersintoronto.org/webinar

See you at the meeting!