BrainArray

BrainArray is an AI-powered decision analysis tool that transforms complex choices into interactive probability trees. Users describe a decision or scenario, and the application leverages Google Gemini to generate a multi-level branching visualization of potential outcomes with associated probabilities.

Currently, the high-performance compute is heavily abstracted and not a part of the core logic. However, a main point of future improvement is HPC integration to have Monte-Carlo simulations for risk analysis and other complex decision analysis tasks.

System Architecture

The application follows a client-server architecture with Auth0 handling authentication and Gemini providing AI inference for decision tree generation.

graph TD
    Client[React Frontend] -->|REST/HTTP| API[Express API Server]
    Client -->|OAuth 2.0| Auth0[Auth0]
    Auth0 -->|JWT Validation| API

    API -->|Store/Retrieve Trees| DB[(MongoDB Atlas)]
    API -->|Generate Tree| Gemini[Google Gemini API]

    subgraph "Frontend Layer"
        Client -->|Render| Tree[Decision Tree Visualizer]
        Client -->|Animate| Motion[Framer Motion]
    end

    subgraph "AI Inference"
        Gemini -->|JSON Response| API
    end

Features

AI-Generated Decision Trees: Describe any decision scenario and receive a structured probability tree with multiple outcome branches
Interactive Visualization: Pan, zoom, and explore decision trees with smooth animations
Persistent Storage: Save and revisit previous decision analyses
Secure Authentication: Auth0 integration for user management and protected API routes

Challenges

Getting the Auth0 set up properly so that it consistently worked and connected with the MongoDB database
Connecting the elements in the decision tree (we used a very simple workaround, and thus it does not look like a normal decision tree would and it is very messy, which is something for us to fix for later)
Making sure the LLM gave correct outputs consistently
Confirming the UI/UX was up to par and intuitive to use