✨ some of my personal and work projects:
- Gandalf: LLM security/prompt injection challenge, see my blog post here
- Running one of the largest prompt-injection red-teams in the world
- Co-created the Gandalf challenge in 2023
- Played by over 750k+ unique players and collected over 23M+ attacks.
- Featured in TechCrunch, Fortune, Sifted, HN, Harvard CS50, Microsoft Build 2024.
- Invited to create versions of Gandalf for Harvard's CS50 and RSAC conference 2024.
- Lakera Guard: building models defending prompt injection attacks and safeguarding agentic AI
- merkle-signature: a fast implementation of merkle signatures with Curve25519 written in Go.
🌟 research:
- AI Security Research: A security benchmark measuring how backbone LLM choice affects AI agent resilience against adversarial attacks, built on 194,331 unique crowdsourced attacks from real human red-teamers on Gandalf: Agent Breaker.
🎓 university projects:
- sphere-tracer: C implementation of a sphere tracer optimized for maximal single-core performance. Using Intel's AVX2 SIMD with object- or ray-parallelization (report).
- acfts: asynchronous consensus-free transaction system, published at ESORICS 2022 (arXiv).