Your result

When completing this training, you will be able to:

• Deploy RAG pipelines on Kubernetes using Helm and the NVIDIA NIM Operator
• Use NVIDIA NIMs for scalable, containerized Large Language Models (LLMs) and embedding models
• Connect, customize, extend, and autoscale application components
• Monitor application performance with Prometheus and Grafana
• Perform load testing and optimize RAG pipelines

Target audience

This training is intended for:

• Machine Learning Engineers who want to deploy RAG applications in production environments
• AI developers who want to implement scalable AI services
• DevOps Engineers who manage Kubernetes environments running AI workloads
• Software Architects who design robust AI architectures
• Technical Leads who want to scale Proof-of-Concepts to enterprise production systems

Approach

The training is practical and interactive. You will work hands-on in a Kubernetes cluster and learn to:

• Set up an infrastructure
• Deploy and configure a RAG pipeline
• Set up monitoring
• Implement autoscaling
• Optimize performance

All components are supported with clear explanations, demonstrations, and realistic lab exercises.

Prerequisits

• Basic knowledge of Kubernetes and container orchestration
• Experience with AI models and LLMs is a plus
• Familiarity with Helm, Prometheus, and Grafana is helpful but not required

Study materials

• Access to interactive labs in a Kubernetes cluster
• Step-by-step instructions through the digital learning environment
• Technical documentation on NIMs, Helm, monitoring tools, and scaling techniques

Certificate

Upon completion of this training, you will receive the certificate of participation ‘NVIDIA: Introduction to Deploying RAG Pipelines for Production at Scale (English)’ by Computrain. 

Content

• Introduction: production-ready RAG pipelines at scale
• Core components: NVIDIA NIMs, Kubernetes, Helm, Prometheus, Grafana
• Kubernetes setup: configuring the interactive learning environment and basic Kubernetes commands
• Deployment: end-to-end deployment of a RAG pipeline with Helm and use of individual NIM services
• Monitoring: applying DCGM, configuring Prometheus and Grafana for performance monitoring
• Autoscaling: implementing horizontal autoscaling (HPA) based on custom metrics
• Load testing: performing load tests and conducting performance analyses