Reginaldo Silva
AI Engineer & Solution Architect | Microsoft AI & Copilot
📋 Biography
I'm an AI Engineer and Solution Architect at Dataside, a Brazilian data and AI consultancy, where I design AI solutions on the Microsoft ecosystem, Azure AI services, Copilot Studio, and Microsoft Fabric, helping enterprise clients move from proof of concept to production. My work includes AI agent architectures, RAG and document intelligence solutions on Azure, and MLOps proofs of value on Microsoft Fabric. I hold Microsoft certifications including Azure Data Engineer Associate, MCSE, and MCSA, plus six Databricks credentials. I'm deeply engaged with the Microsoft community in Brazil: I spoke at the Microsoft AI Tour, SQL Saturday, and Azure Bootcamp, co-lead the SQL Vale chapter in São José dos Campos, and publish technical content in Brazilian Portuguese at datainaction.dev, covering AI agents, Copilot, and Azure. My mission is to help the Brazilian community build real, production-grade AI solutions with Microsoft technologies.
✨ High-Impact Contributions 5
Explaining the new Github Copilot Harness and Standard one.
What we'll cover in this post: The big picture: the 6 blocks on the agent screen Instructions: the personality and the rules Knowledge: where the answers come from Tools: what the agent does (and what Work IQ is) Triggers and Agents: proactivity and multi-agent Topics: flows drawn by hand
What we'll see in this post: What we're going to build (the DataDay case) Step 1: create the agent and write the Instructions Step 2: plug SharePoint in as Knowledge Step 3: Tools and Work IQ (what the screenshot shows underneath) Step 4: test the agent and read the grounding A Web Search gotcha Bonus: a scheduled Trigger to make the agent proactive
What we'll cover in this post: Why the name "Copilot" turned into alphabet soup The map: what you only use off the shelf vs. what you actually build Copilot Studio vs Microsoft Foundry: low-code vs pro-code How to choose between the two (and why it's not "either/or")
AI Agent Simulator is an open source educational visualizer that exposes the complete lifecycle of an agentic AI request, animated in real time. Most engineers now build with agents without ever seeing what happens inside a single request. The simulator makes it visible end to end: frontend to backend to agent to LLM to database, the ReAct loop as the agent reasons and calls tools, real RAG including chunking, embeddings, vector search, reranking and hybrid search, MCP tool calls the agent decides to make on its own, and the context window being assembled token by token. Nothing is mocked, every stage executes real code against real services, so what the user watches is the actual behaviour of a production-shaped agent rather than an animated diagram. I use it as teaching material in talks to explain agent internals to engineers starting in AI Engineering, i use to explain RAG in Foundry and Copilot Studio too.