Alhamza Ibrahim

• Platform-agnostic: every agent is exposed over the A2A protocol and can be reached from Microsoft 365 Copilot, SAP Joule or a custom UI.
• Built generically, so a customer picks the agents they need and most of the work is already done. What remains is adapting them to that customer's system.
• The agents carry complex workflows from start to finish: document extraction, data enrichment, agent-to-agent communication, human-in-the-loop review and the final approval.

LangChain / Azure AI Foundry / SAP BTP / A2A / FastAPI / Python / TypeScript

2nd Runner-Up · SAP Hack2Build 2025

• One assistant over everything the plant knows: unstructured manuals, PDFs and SharePoint pages, next to structured ERP data reached through an MCP server.
• Multimodal in both directions. Workers speak, type or show it a photo; it understands speech and images and answers in text or voice.
• Runs as a scalable microservice, so one deployment serves the whole production line.

Azure OpenAI / FastAPI / FastMCP / RAG / Azure Speech / Vision

• Two kinds of knowledge behind one chat: unstructured FAQs and manuals, and a structured database of the vehicles in stock.
• A routing layer reads the intent of each message and sends it either to document retrieval or to a text-to-SQL agent that queries live vehicle data.

GPT / LangChain / FastAPI / Pinecone / BERT / Azure DevOps CI/CD

• Approval workflows in Teams built with Adaptive Cards and the Microsoft 365 Agents SDK, so the decision stays with a human.
• Multi-agent orchestration across Azure AI Foundry and Copilot Studio, connected through MCP and A2A.
• Event-driven triggers with Azure Logic Apps, and enterprise knowledge retrieval over SharePoint and Azure AI Search.

M365 Agents SDK / Copilot Studio / Azure AI Foundry / Logic Apps / MCP / A2A