PinnedInAgentic AI VantagebyRoshan Kumar·Jul 7The Cognitive Governance Matrix: 25 Design Patterns for Production-Grade Agentic AISimple prompt engineering is dead. Here is the definitive 25-pattern architectural blueprint to safely scale autonomous AI agents.
PinnedInArt of Product ManagementbyRoshan Kumar·Oct 9, 2025The Art of the ‘Generative No’Master the Generative No: Reject the solution, not the problem. This simple shift turns stakeholders into partners and builds better…
InAgentic AI VantagebyRoshan Kumar·3d agoGoogle Cloud Agentic Architect Certification Exam Prep (Part 5 of 5) — Domain 5Master Domain 5 of the Google Cloud Professional Agentic Architect certification with architectural questions on Agent Gateway, Model Armor…
InAgentic AI VantagebyRoshan Kumar·4d agoGoogle Cloud Professional Agentic Architect Certification Exam Prep (Part 4 of 5) — Domain 4Master Domain 4 of the Google Cloud Professional Agentic Architect certification with architectural questions on trajectory match modes…
InAgentic AI VantagebyRoshan Kumar·5d agoGoogle Cloud Agentic Architect Certification Exam Prep (Part 3 of 5) — Domain 3Master Domain 3 of the Google Cloud Professional Agentic Architect certification with architectural questions on ADK LlmAgent…
InAgentic AI VantagebyRoshan Kumar·6d agoGoogle Cloud Agentic Architect Certification Exam Prep (Part 2 of 5) — Domain 2Master Domain 2 of the Google Cloud Professional Agentic Architect certification with architectural questions on Google Antigravity (agy)…
InAgentic AI VantagebyRoshan Kumar·Nov 12Google Cloud Agentic Architect Certification Exam Prep (Part 1 of 5) — Domain 1Master Domain 1 of the Google Cloud Professional Agentic Architect certification with rapid-fire architectural questions on Workflow…
InAgentic AI VantagebyRoshan Kumar·Nov 6ADK Agent Stops Calling Tools the Moment You Enforce Strict JSONWhy enabling strict JSON output_schema disables ADK tool calls—and how a 2-step SequentialAgent split fixes it while cutting tokens.
InAgentic AI VantagebyRoshan Kumar·Nov 3Why Workload Identity alone fails on User-Specific ToolsHow a single shared service account lets one user hijack another user’s portfolio — and the 3-layer zero-trust architecture that stops it…
InAgentic AI VantagebyRoshan Kumar·Nov 1Deterministic ADK Callbacks & Human-in-the-Loop (HITL) on Google CloudPart 3 of the Agentic Security Trilogy: Mastering before_model_callback, before_tool_callback, and Tool-Level vs. Conversation-Level HITL