null Chennai HUMLA Session - Hands-on AI Security

​null HUMLA Session - LLM Top 10, Guardrails, Building an Agent and Agentic AI security Issues

​Please update this google form to select the participants :(MANDATORY)

​https://form6s.gle/12GJvxgzRW1Ykshj

​Speakers :

​Meenakshi Ganesh and Dilip Sai

​https://www.linkedin.com/in/meenakshi-ganesh-3aa2a5258
https://www.linkedin.com/in/dilip-sai-a0b3861b4

​Agenda:

Part 1 – Intro + Setup (1 hr) Quick intro to LLMs, what they are. Show how to deploy a small model on Colab. Interact with it using LangChain (basic example).

Part 2 – OWASP LLM Top 10 (3 hrs) Short explanation with a hands on demo for each category

Part 3 – Guardrails (1 hr) Show how guardrails work on the inside and how it impacts model answers. Followed by a demo of the same.

Part 4 – RAG (1 hr) Quick overview of RAG, vector DBs and RAG pentesting.

Part 5 – Agents + Security (1 hr) How to build an agent. Agentic AI security issues.

Q&A

​Requirements : 16 Gigs of RAM

Beginner level coding experience

VS Code + Python configured locally on windows

Sign up here for workshop pre-reqs: Google Colab account : https://colab.research.google.com/

​NVIDIA API key: https://build.nvidia.com/

​PortSwigger account : https://portswigger.net/web-security

​Crucible Labs account : https://platform.dreadnode.io/

​What not to bring? Work laptops with too many restrictions