Overview
A serverless machine learning inference pipeline built on AWS infrastructure with a Telegram Bot interface for on-demand model predictions.
Technical Stack
- Cloud: AWS Lambda, API Gateway, S3, DynamoDB
- Infrastructure: Terraform IaC for reproducible deployments
- Development: LocalStack for local testing
- Runtime: Dockerized Python for portable ML deployments
- Interface: Telegram Bot API for user interaction
Key Features
- Serverless Architecture: Pay-per-use model with AWS Lambda for cost-efficient inference
- Infrastructure as Code: Version-controlled, reproducible cloud infrastructure with Terraform
- State Management: DynamoDB for lightweight request logging and state management
- Portable Deployments: Docker containers for consistent ML model execution
- User-Friendly Interface: Telegram bot for easy, on-demand model inference requests
Architecture
- User sends request via Telegram Bot
- API Gateway receives and routes the request
- Lambda function processes input and runs inference
- Results stored in DynamoDB and returned to user
- S3 for model artifacts and data storage