Kai Greshake

Senior AI ML Security Architect at NVIDIA

Saarbrücken, Saarland, Germany
email-iconphone-icongithub-logolinkedin-logotwitter-logostackoverflow-logofacebook-logo
Join Prog.AI to see contacts
email-iconphone-icongithub-logolinkedin-logotwitter-logostackoverflow-logofacebook-logo
Join Prog.AI to see contacts

Summary

🤩
Rockstar
Kai Greshake is a Senior AI/ML Security Architect with a decade of experience, currently focused on product security and offensive AI/ML research at NVIDIA. Based in Saarbrücken, Germany, he blends hands-on security engineering with research-driven exploit development to identify real-world risks in app-integrated LLMs. His open-source work includes the llm-security project, which demonstrates prompt-injection and data-exfiltration attacks and builds “poisoned” tool demos to show how LLMs can be manipulated. He also contributes to systems-level Rust projects like i3status-rust, applying functional idioms and low-level optimizations to resource-friendly tooling. Known for turning sophisticated attack techniques into reproducible research and defensive guidance, he writes and shares insights on his technical blog. Colleagues value his mix of offensive creativity and practical engineering that helps harden AI products before they ship.
code9 years of coding experience
github-logo-circle

Github Skills (16)

it-security10
security-testing10
vulnerabilities10
exploit10
rust10
api10
python10
api-design10
langchain10
vulnerability10
security10
develop9
system-design8
command-line-interface7
dockers5

Programming languages (9)

JavaShellRustCJavaScriptHTMLJupyter NotebookPython

Github contributions (5)

github-logo-circle
greshake/llm-security

Feb 2023 - Feb 2023

New ways of breaking app-integrated LLMs
Role in this project:
userSecurity Engineer
Contributions:57 commits, 2 PRs, 34 pushes in 9 days
Contributions summary:Kai primarily worked on enhancing the security of LLMs integrated into applications. Their contributions involved developing tools and implementing various attack scenarios, notably focusing on prompt injection and data exfiltration techniques within a "bong chat" application. They designed and implemented "poisoned" tools such as search, website viewing, and URL retrieval, incorporating features to detect and respond to specific triggers and payloads. Furthermore, the user created and refined demonstrations showcasing how vulnerabilities can be exploited to gain control of an LLM and extract sensitive user information.
greshake/i3status-rust

Apr 2017 - Oct 2020

Very resourcefriendly and feature-rich replacement for i3status, written in pure Rust
Role in this project:
userFull-stack Developer
Contributions:1 release, 169 commits, 72 PRs in 3 years 7 months
Contributions summary:Kai's contributions primarily involve modifying the Rust code for the i3status-rust project. They have worked on updating the main file and adding comments and TODOs. Moreover, the user is modifying blocks (e.g., battery, music, cpu, memory) and implementing features like a toggle block, and adding sound and temperature functionalities. These changes include incorporating functional programming styles and using Rust's conventions.
rustfeature-richpure-rusti3statusreplacement
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.
Request Free Trial
Kai Greshake - Senior AI ML Security Architect at NVIDIA