The best way to make the modules stick is to practice them. These five projects take you through the full cycle: break a model with prompt injection, red team a chatbot with garak, protect the AI supply chain, add guardrails around an LLM, and finally set up an AI governance program.
Each one is a video you can follow along with, so you can pause, try it yourself and adapt it to your own ideas. Only test systems you own or have permission to test. Pick the project that matches the module you have just finished and come back for the rest.
Work through a series of gamified labs that use prompt injection to get an AI to give up a secret, with defenses that get stricter at each level.
Install garak, an LLM vulnerability scanner, and use it to test your own REST-based chatbot for weaknesses.
See how to protect LLM workloads, pipelines, and model artifacts end to end using HashiCorp Vault.
Build a chatbot in Python and put NVIDIA NeMo Guardrails around it to control topics and behavior.
Follow a five-step approach to governing high-risk AI systems, from intake through ongoing oversight, using a real use case.