Deep DiveAI Engineering
Building Domain-Specific Tools -- Giving Agents Real Capabilities
How to build production-quality agent tools that query databases, validate business rules, and automatically scope data to the current user.
Read article
Journal
Solutions Architect specializing in Azure, AI, and enterprise cloud-native development. Writing about architecture, AI agents, and building things that scale.
Deep DiveAI Engineering
Multi-Agent Architecture: Orchestration and the A2A Protocol
How to build an orchestrator that routes requests to specialist agents, and the A2A protocol that makes agent-to-agent communication standardized, secure, and framework-independent.
Deep DiveAI Engineering
Observability -- Tracing Multi-Agent Workflows with OpenTelemetry
How to instrument multi-agent systems with OpenTelemetry -- auto-instrumented traces across agents, LLM calls, and database queries, visualized in .NET Aspire Dashboard.
Deep DiveAI Engineering
Frontend: Rich Cards and Streaming Responses
Transform agent text responses into interactive product cards and order cards, then add token-by-token streaming via SSE so the UI feels instant -- all without changing the backend agents.
Read article
Deep DiveAI Engineering
Production Readiness: Auth, RBAC, and Deployment
JWT authentication, role-based access control, user-scoped data isolation for a multi-agent system, then Docker Compose architecture with a one-command startup for all 11 services.
Deep DiveAI Engineering
Agent Memory -- Remembering Across Conversations
Give your agents persistent memory -- store user preferences, recall past interactions, and personalize responses across conversations.
Deep DiveAI Engineering
Evaluating Agent Quality -- Testing What You Cannot Unit Test
Build a repeatable evaluation pipeline for multi-agent correctness -- golden datasets, automated scoring, and CI/CD integration.
Deep DiveAI Engineering
MCP Integration -- Connecting AI Agents to the Tool Ecosystem
Replace hand-coded tools with Model Context Protocol servers -- standardized tool discovery and execution for any agent framework.
Deep DiveAI Engineering
Graph-Based Workflows -- Beyond Simple Orchestration
Move from LLM-driven routing to deterministic graph-based workflows -- sequential pipelines, parallel execution, conditional branching, and state management.
Deep DiveAI Engineering
MAF v1 — Sessions (Python + .NET)
Serialize an AgentSession to JSON, persist it, reload in a fresh process, and have the agent pick up exactly where it left off.
Deep DiveAI Engineering
MAF v1 — Streaming and Multi-turn (Python + .NET)
Stream tokens as they arrive and reuse a session across turns so the LLM sees the full conversation. ~60 lines of code per language.
Deep DiveAI Engineering
MAF v1 — Adding Tools (Python + .NET)
Give an agent a callable function, let the LLM decide when to use it, and compare Python and .NET tool loops side by side.