I develop high-level decision-support systems that move beyond simple chatbots. MASAS is a "Headless" AI engine that simulates a professional board of directors to stress-test business strategies and eliminate cognitive bias. This project demonstrates my ability to build complex, recursive AI loops that deliver structured, executive-ready intelligence.

The Architecture:

Built on n8n, the system uses a central Orchestrator (The Chairperson) to coordinate a multi-stage debate between specialized sub-workflows:

• Visionary & Skeptic: For adversarial reasoning and risk discovery.
• Operator & Analyst: For feasibility and data-driven grounding.
• Ethicist: For compliance and brand alignment.

https://youtu.be/a-CWk5d_5M4

Key Technical Highlights:

• Sub-workflow Partitioning: Each agent (Visionary, Skeptic, etc.) is isolated in its own sub-workflow. This ensures modularity, making it easy to update specific "agent personalities" without breaking the main engine.
• Adversarial AI Logic: Implements a multi-round "Challenge-Response" cycle where the Skeptic identifies flaws in the Visionary's proposal, forcing the AI to provide more robust and realistic advice.
• Contextual Routing: Uses intelligent routing to determine which agents need to be "called to the board" based on the user's specific query type.
• Structured Output Synthesis: Consolidates diverse perspectives from 5+ LLM calls into a single, cohesive Markdown/JSON report, ready for API consumption or frontend rendering.
• Prompt Engineering for Personas: Deeply researched system prompts that define specific professional constraints, ensuring the AI stays in character and provides diverse viewpoints.