Harish Ramadoss ,
Harish Ramadoss has several years of expertise in Product Security, Red Teaming, and Security Research.
Previously, he was a Principal at Trustwave Spiderlabs, where he led their Application Security efforts. He joined Rippling as a founding member of the Security Engineering team and leads their AI Security and Appsec efforts.
Harish built DejaVu, an open-source deception platform. He has presented at Black Hat, DEFCON, HITB, and other conferences globally.
Training: Practical AI Security - Go Beyond Theory: Build, Break, and Defend
Before you can secure or break AI applications, you need to understand how they’re built.
Build
This hands-on training starts with first principles and an engineering mindset. You’ll learn how real GenAI applications are built - from interacting with LLM APIs to working with embeddings, VectorDBs, RAG pipelines, agentic systems, MCPs, LangSmith and essential tooling. Through guided labs, you’ll explore these components deeply and ultimately build your own RAG based threat-modeling agent.
Break
Next, we shift into offensive security by attacking real applications in our labs. You’ll practice classic and indirect prompt injection, sensitive data disclosure, and agentic-architecture attacks. We also cover MCP exploitation by building custom MCP servers, and explore model backdoors using real-world examples. You’ll learn to reason about threats across the entire AI stack.
Defend
Finally, we focus on defense; implementing guardrails, using MCP gateways for observability, fixing vulnerabilities shown in earlier labs, and applying agentic security patterns. We also cover AI security tooling and how to integrate it into the SDLC.