Justin Shenk

Justin works on AI safety and helps people understand how models work and how to get involved. He researches interpretability for LLMs and contracted for Redwood Research. He facilitates AGI Strategy and Frontier AI Governance courses at BlueDot Impact and organizes AI Salon Berlin.

He is studying how time is represented in LLMs (temporal-awareness), evaluating models for information integrity (civicevals.org), and mentoring AI Safety Camp and SPAR projects.

Previously, he co-founded VisioLab, a computer vision startup doing food recognition for self-checkout, where he served as CTO. He received his Ph.D. at the Donders Institute (Radboud University, the Netherlands), where he developed Traja, a Python library for analyzing behavioral trajectories. His M.Sc. thesis at the University of Osnabrück focused on identifying interpretability signals from neural network activation covariance.