Z360 v1 tried to solve everything for everyone. It was unsustainable. We pivoted to AI agents. My role evolved from shipping features to designing platforms. I led a >5 person R&D lab interviewing all applicants, and architected a multi-repo agent platform.
What I Built
Led a >5 person AI R&D cycle, delivering LangGraph orchestration architecture and Python-Laravel interop via HTTP/MCP
Designed voice AI infrastructure using LiveKit + LangGraph adapters, achieving realtime intelligence
Led development of a highly customizable and intelligent agentic customer support system with seamless integration across SMS, Email, and VoIP channels
Championed trunk-based development, Tests, Lints and Type Checks. v2 staged in 2 months vs v1's 8 months
Key Wins
Established a culture of quality over quantity, achieving simple, maintainable code with high long-term returns
Implemented strict processes over intuition (e.g., backend testing, frontend linting, type-checking), which significantly improved platform stability and reduced bugs before code was ever shipped.
Gained first hand experience designing AI-agent architectures, integrating LLMs with real-time communication channels, and building scalable serverless infrastructures.
The Lesson
Foundations determine your fate: strong ones scale, weak ones pivot. Code is risky without guardrails; with them, it becomes an asset.