decodingai-magazine/llm-twin-course

Covers the complete MLOps pipeline across four microservices: data collection via web crawlers and CDC patterns into MongoDB, real-time feature engineering with Bytewax streaming into Qdrant vectors, fine-tuning with LoRA/QLoRA tracked in Comet ML, and inference deployment on AWS SageMaker with RAG enhancement and prompt monitoring via Opik. Integrates with Hugging Face model registry, Redis/Qdrant vector databases, RabbitMQ for event streaming, and AWS Lambda for serverless data collection.

4,297 stars. No commits in the last 6 months.