## Role Overview As a Software Engineer at PostQode, I architect and ship production features for an AI Agent Marketplace platform — enabling users to discover, purchase, and deploy AI agents into their Azure subscriptions through a resilient multi-stage orchestration pipeline with real-time deployment tracking. ## Key Deliverables - **AI Agent Marketplace with Azure-Native Deployments** — Built the full marketplace using NestJS, Next.js, Azure ARM templates, and Socket.IO. Users can discover, purchase, and deploy AI agents directly into their Azure subscriptions. The system implements a 10-stage orchestration pipeline with recoverable failure handling, payment workflows, agent catalog management, and secure post-purchase Azure credential handoff. - **Agentic Chatbot (Tool-Calling Architecture)** — Built an AI assistant on an agentic tool-calling loop where the LLM autonomously selects and invokes RBAC-gated tools for workspace management, license verification, Azure operations, and agent purchases. Dual-modality: text via SSE streaming and voice via Vapi integration with real-time LLM response rendering. - **Git Tracker** — Owned end-to-end delivery of a real-time Git status tracker using Node.js child processes and file watchers. Reduced sync latency 93% (3s → 200ms) with 100% accuracy in nested change detection across large monorepos. ## Technical Approach Every feature starts with understanding the failure modes — the Azure deployment pipeline handles partial failures with recoverable state, the tool-calling chatbot validates RBAC before invoking any tool, and the Git tracker uses incremental diff computation to avoid full-tree rescans.