AI Agent
From tool calling, MCP, memory and RAG to workflows and production deployment.
When AI Agents Need Human Approval: Designing Automation by Risk
Human approval is not an automation failure; it puts high-impact actions at the right accountability point.
Evaluating RAG Before Production: Accuracy, Citations and Freshness
Production RAG needs evaluation for retrieval, citations, permissions, freshness and refusal quality rather than merely loading documents into a vector store.
Reliable AI Agent Tool Calling: From Calling Tools to Delivering Outcomes
Reliable tool use needs contracts, validation, timeouts, idempotency and recovery instead of a model simply receiving a list of functions.
When AI Agents Need Human Approval: Designing Automation by Risk
Human approval is not an automation failure; it puts high-impact actions at the right accountability point.
Evaluating RAG Before Production: Accuracy, Citations and Freshness
Production RAG needs evaluation for retrieval, citations, permissions, freshness and refusal quality rather than merely loading documents into a vector store.
Reliable AI Agent Tool Calling: From Calling Tools to Delivering Outcomes
Reliable tool use needs contracts, validation, timeouts, idempotency and recovery instead of a model simply receiving a list of functions.
MCP and Governed AI Tools: Build the Gateway Before the Agent
How standard tool connections can become reliable enterprise infrastructure when identity, consent and audit are designed first.
Agent Memory in Production: Keep Useful Context, Remove the Rest
A practical guide to memory layers, retention, consolidation and deletion for long-running AI agents.
AI Agents in Production: Autonomy Needs Approval, Memory and Audit
A practical framework for taking tool-using agents from a demo to a safe and measurable production workflow.