AI Engine

GPU infrastructure management system for deploying isolated GPU instances on Crusoe Cloud with a web-based dashboard.

Overview

AI Engine provides a complete solution for managing GPU resources for teams, featuring:

• Web-based dashboard for team and GPU allocation management
• Automated Terraform deployment to Crusoe Cloud
• Isolated GPU environments with persistent storage
• Real-time resource tracking and bin packing visualization

Architecture

Frontend (Next.js)

• Modern React-based web dashboard
• Real-time GPU quota tracking
• Visual bin packing algorithm demonstration
• Dark mode support
• Located in /frontend

Backend (Terraform)

• Infrastructure as Code for Crusoe Cloud
• Automated VM provisioning with GPU allocation
• SSH key generation and management
• Located in root directory (main.tf)

API Routes

• /api/teams - Team CRUD operations
• /api/deploy - Trigger infrastructure deployment

Prerequisites

• Node.js >= 18
• Terraform >= 1.0
• Python 3 >= 3.7
• Crusoe Cloud account with API credentials

Quick Start

1. Configure Crusoe Cloud

# Create ~/.crusoe/config
[default]
access_key_id = your_access_key
secret_access_key = your_secret_key

2. Install Dependencies

# Frontend
cd frontend
npm install

# Python scripts
pip3 install json argparse datetime pathlib

3. Run the Application

# Start the web dashboard
cd frontend
npm run dev

Access the dashboard at http://localhost:3000

4. Deploy Infrastructure

Use the "Deploy" button in the web interface or run:

./scripts/deploy.sh

Features

Team Management

• Add teams with custom GPU allocations (1, 2, 4, 8, or 10 GPUs)
• Specify team members and custom ports
• Real-time GPU quota tracking

Infrastructure

• Each team gets dedicated NVIDIA L40S GPUs
• Pre-configured with development tools
• Persistent storage across restarts
• SSH and Jupyter notebook access

Resource Visualization

• Bin packing algorithm shows optimal VM placement
• Visual representation of host utilization
• Real-time allocation feedback

Project Structure

AI-Engine/
├── frontend/              # Next.js web application
│   ├── app/              # App router pages and API routes
│   ├── components/       # UI components
│   └── public/           # Static assets
├── data/                 # Team configuration data
├── scripts/              # Deployment and utility scripts
├── main.tf              # Terraform infrastructure
└── README.md

Access Information

After deployment:

• SSH keys: keys/[team]_private_key
• Connection scripts: outputs/connect_[team].sh
• Jupyter notebooks: http://[team_ip]:8888