I Boardroom OS: A Private Board of Directors That Runs on a 4GB GPU

AI Boardroom OS is an innovative multi-agent business decision simulator that functions entirely on a local GPU with 4GB of memory. The platform simulates a virtual boardroom of 10+ executive personas, including CFO, CTO, CMO, and more. Users can describe a business decision, upload supporting documents, and the AI agents analyze the situation, debate it, and present a final recommendation to a CEO agent who synthesizes the best course of action.

The simulator is designed for solo founders, small agencies, and consultancies who want a structured second opinion on pricing, hiring, or go-to-market strategies, but cannot afford to hire an advisory board or share confidential information with third-party APIs. The entire system runs on the user's own machine, providing a secure and private environment for decision-making.

The platform employs a four-stage workflow: executive analysis, debate, CEO synthesis, and decision validation. To optimize performance on small models like Qwen2.5-3B-Instruct (4-bit NF4 quantized), the developers implemented several small-model tricks such as JSON-schema-constrained outputs, role-specific grounding hints, and compressed "cards" passed between stages to keep the context small.

The AI Boardroom OS is built on a robust stack of Python, PyTorch, Hugging Face transformers, FastAPI backend, and Streamlit multipage frontend. The platform offers two intelligence modes, three response formats, and allows for manual or automatic agent selection. The code is available on GitHub under the username krushnakodgirwar.