Explore the hidden dangers of prompt injection in Large Language Models (LLMs). This lab reveals how attackers manipulate LLMs to disclose private information and behave in ways that they were not intended to. Discover the intricacies of direct and indirect prompt injection and learn to implement effective guardrails.

This hands-on lab provides a comprehensive introduction to NVIDIA Run:ai, a powerful platform for managing AI workloads on Kubernetes. Designed for AI practitioners, data scientists, and researchers, this lab will guide you through the core concepts and practical applications of Run:ai's workload management system.

In this lab you will explore an NVIDIA Data Flywheel running on Cisco Secure AI Factory hardware with NVIDIA in WWT's AI Proving Ground. Built on the NVIDIA NeMo microservices platform, the flywheel automatically turns production traffic into evaluation and fine-tuning datasets, LoRA fine-tunes candidate models with NeMo Customizer, scores them against the incumbent with NeMo Evaluator, and deploys the winner as an NVIDIA NIM. Using a clinical coding application as the example workload, you will watch a 1-billion-parameter model close the accuracy gap on a 70-billion-parameter model at a fraction of the inference cost.

This lab will go into the basics of Retrieval Augmented Generation (RAG) through hands on access to a dedicated environment.