From Distillation to Reinforcement Fine-Tuning: Hill-Climbing in Microsoft Foundry
Improving a deployable model package one measured step at a time across quality, latency and cost—comparing SFT distillation with reinforcement fine-tuning.
Microsoft Tech Community
Articles I have published on the Microsoft Foundry Blog on Microsoft Tech Community—hands-on architectures for agentic AI, MCP integration, computer-use agents, real-time voice agents and model fine-tuning. 16 articles and 30,000+ reads.
Driving Stop-and-Go Business Processes to Closure with Foundry Hosted Agents
Building agents that track work across days or weeks—keeping state so long-running business processes actually reach closure.
Interactive AI Avatars: Building Voice Agents with Azure Voice Live API
Speech-to-speech AI avatars built on the newly GA Azure Voice Live API.
Building Enterprise Voice-Enabled AI Agents with Azure Voice Live API
An end-to-end voice agent covering product search, order management, automated shipment creation and analytics.
Autonomous Visual Studio Code Desktop Automation using Computer Use Agent & PyAutoGUI
Combining Computer Use Agent models with PyAutoGUI to drive fully autonomous VS Code workflows.
Automate a Multi-Step Business Process using Turnkey MCP and Logic App Integration in AI Foundry
Procure-to-Pay anomaly detection with Azure AI Foundry Agent Service—analysing invoice images for procurement anomalies and compliance.
AI Automation in Azure Foundry through Turnkey MCP Integration and Computer Use Agent Models
A fashion-trends discovery scenario combining Computer Use models with Playwright browser automation and MCP.
Dynamic Tool Discovery: Azure AI Agent Service + MCP Server Integration
Connecting Azure AI Agent Service to Model Context Protocol servers for dynamic tool discovery—before turnkey support existed.
Using the CUA Model in Azure OpenAI for Procure-to-Pay Automation
Solution architecture for P2P automation using the Computer Use Agent model and the Responses API.
Agentic P2P Automation: Harnessing the Power of OpenAI's Responses API
Turning an error-prone, manual Procure-to-Pay process into an agentic workflow with the Responses API.
AutoGen & OpenAI Assistants API Powered Collaborative Report Generation
Multi-agent collaboration with AutoGen to generate monthly sales-performance reports from cloud-stored data.
From Vector Databases to Integrated Vector Databases: Revolutionizing AI-Powered Search
Why integrated vector databases change the game for semantic and vector search behind AI assistants.
Exploring Azure AI Agent Service: A Leap in Conversational AI
What Azure AI Agent Service adds beyond the Assistants API—state management, automated tool actions and more.
Working with the Realtime API of gpt-4o in Python
Connecting to the Realtime API and handling audio conversations, text conversations and tool calling.
Building an AI Assistant Using the gpt-4o Audio-Preview API
Building an assistant on the gpt-4o Audio-Preview API—and how it differs from the Realtime API.
Multimodal Video Search Powered by Video Retrieval in Azure
Multimodal search across video content—from training material to safety monitoring—with Azure Video Retrieval.