🚀 Cost-Efficient AI Development Platform Using Spot Instances

스팟 인스턴스를 활용한 비용 효율적인 AI 개발 플랫폼

Callisto is an open-source platform that combines serverless computing and spot instances to provide a highly cost-effective AI development environment. It automatically provisions and terminates resources, supports secure authentication, and minimizes idle resource costs — ideal for research, education, and lightweight production use cases.

Callisto는 서버리스 컴퓨팅과 스팟 인스턴스를 결합하여 비용 효율적인 AI 개발 환경을 제공하는 오픈소스 플랫폼입니다. 자원을 자동으로 생성·회수하고, 안전한 인증을 지원하며, 유휴 리소스의 비용을 최소화합니다.

---

✨ Features

• 🔐 Serverless authentication with JWT
  Stateless, secure user authentication using JSON Web Tokens (JWT).

• ☁️ Dynamic spot instance provisioning
  Uses AWS price-capacity-optimized strategy to launch cost-effective and stable spot instances.

• 💤 Idle session auto-termination
  Automatically shuts down unused environments to prevent unnecessary charges.

• 📉 Verified cost savings
  Up to 77.44% reduction in real-world classroom deployments.

🏗 Architecture

• Frontend: React-based UI for login and Jupyter session requests
• Backend: Stateless API via AWS Lambda and API Gateway
• Compute: EKS + Karpenter for spot-based auto-scaling

📦 How to deploy

• You need change the contents of variables.tf file.

• Run the command below :

  terraform init
  terraform apply --auto-approve

📊 Cost Comparison

Platform	Monthly Cost (Est.)	Savings
AWS SageMaker	$2,168.00	-
Azure ML Notebook	$1,780.00	-
GCP Vertex AI Workbench	$1,650.00	-
Callisto	$516.00	76.2%↓

In a real university class with 40 students, the total cost was only $34.43 over 50 days.

🤝 Contributing

• Fork the repo and create a PR from dev branch
• Submit issues for bugs or feature requests

📄 License

This project is licensed under the terms of the GNU General Public License v3.0.