Enterprise AI & Power Platform Engineering — architectures, experiments, and patterns that survive production.
The open-source home of powerailabs.dev — an engineering notebook in public for architects and lead engineers shipping Copilot, agents, and RAG on the Microsoft AI stack. AI · Copilot · Power Platform · Dynamics 365 · Azure AI.
powerailabs.dev — architecture deep-dives, lessons from production, and field notes from the Microsoft AI stack. Cross-posted as a LinkedIn newsletter — follow @mishraraghav to get it in your feed.
Latest posts
- Authentication patterns for Microsoft Foundry — beyond DefaultAzureCredential — managed identities, workload identity federation, OBO, and per-environment role assignment.
- Starting an Azure Foundry project — the getting-started guide nobody wrote — five decisions, the day-one auth pattern, a first-sprint checklist, and three traps.
- Why I built 6 agents instead of 1 mega-agent — lessons from TrafficIQ — routing, handoff, MECE tools, and grounding lessons from production.
- 🥈 1st Runner Up — Microsoft Dynamics 365 Customer Insights ITeS Hackathon — financial-crime intelligent automation on Azure ML, Customer Insights, and D365 case management.
- 🏆 Winning Best Use of Microsoft Foundry at AI Dev Days Hackathon — TrafficIQ — the story, the stack, and the 6-agent / 56-tool architecture.
- Building LocalRAG — a fully local AI document search — Retrieval-Augmented Generation with no cloud LLM and no data egress.
Topics: AI Agents & RAG · AI Engineering · Architecture · Azure AI · Microsoft Foundry · Lessons from Production · Dynamics 365 · Power Platform
- 🤖 AI & machine learning — intelligent, local-first AI systems.
- ⚡ Power Platform & automation — production workflows with Power Apps, Dataverse, Power Automate.
- ☁️ Cloud & on-prem AI — scalable AI solutions, on-prem and on Azure.
- 📚 RAG & LLMs — Retrieval-Augmented Generation, local LLMs, and grounding patterns for the enterprise.
- 🧩 Agents & MCP — Foundry Agent Service, Copilot Studio, multi-agent routing, and Model Context Protocol tools.
In this organisation
- Local RAG — fully local Retrieval-Augmented Generation. No cloud LLM, no data egress. → Project page
Related work (under @raghavmishrad365)
- TrafficIQ — 6 specialist agents, 56 function-calling tools, 3-tier routing (sticky → keyword → LLM). Built on Foundry Agent Service (GPT-4.1 / GPT-4.1-mini), Azure Maps, D365 F&O via MCP, Azure IoT Hub, and Dataverse. 🏆 Best Use of Microsoft Foundry — AI Dev Days Hackathon 2026 (selected from 401 submissions / 2,041 registrants). → Project page
- Make Life Easy — accessibility-first Power Platform solution. 🏆 Global winner, Best DEI Use Case — Microsoft HackTogether: Power Platform Global AI Hack 2023. → Project page
- Learn Fine-Tuning: A Hands-On Course — 19-lesson hands-on course on fine-tuning language models with PyTorch and Hugging Face. Covers foundations, transfer learning, SFT, LoRA / QLoRA, RLHF, DPO, and production deployment.
- Learn LLMOps: A Hands-On Course — 27 lessons + 3 projects on building, deploying, and managing LLM systems in production. Covers RAG, fine-tuning, evaluation, model serving, security, cost optimisation, and AI dev infrastructure.
- 🏆 Best Use of Microsoft Foundry — AI Dev Days Hackathon 2026 (TrafficIQ).
- 🏆 Best DEI Use Case — Microsoft HackTogether: Power Platform Global AI Hack 2023 (Make Life Easy, global winner).
- 🥈 1st Runner Up — Microsoft Dynamics 365 Customer Insights ITeS Hackathon 2021 (financial-crime intelligent automation).
- AI / ML: Microsoft Foundry, Foundry Agent Service, GPT-4.1, Ollama, FAISS, local LLMs, PyTorch, Hugging Face Transformers, PEFT, TRL.
- Microsoft cloud: Azure AI, Azure AI Search, Azure Maps, Azure IoT Hub, Dataverse, Dynamics 365 F&O, Copilot Studio, Microsoft 365 Copilot.
- Integration: Model Context Protocol (MCP), Power Automate, function-calling tools.
- Web & API: FastAPI, React, Azure.
- Data processing: OCR, PDF / DOCX parsing, embedding models.
- 📝 Blog: powerailabs.dev
- 💼 LinkedIn: linkedin.com/in/mishraraghav (newsletter cross-posts here)
- 💻 GitHub: Follow this organisation for updates.
PowerAI Labs is a space for engineering AI-powered solutions on the Microsoft stack — built by Raghav Mishra, Senior Product Architect (AI & Power Platform).