Eitan Sprejer

Eitan Sprejer is the Founding Director of BAISH (Buenos Aires AI Safety Hub), which he scaled to 150+ members and a seven-person organising team as of mid-2026, and a Course Facilitator at BlueDot Impact across six technical AI safety cohorts. He co-founded LANAIS (Latin American Network for AI Safety), a regional network launched in June 2025 to connect researchers, students, and organisers across Latin America. Sprejer holds a Licenciatura in Physical Sciences from the Universidad de Buenos Aires (April 2025, GPA 9.22/10) and is a MARS Fellow at the Cambridge AI Safety Hub, researching automated validity checks for LLM benchmarks. His research focuses on chain-of-thought interpretability, LLM evaluation, and mechanistic interpretability; he is first author of "Approximating Human Preferences Using a Multi-Judge Learned System" (NeurIPS 2025 LatinX in AI and Reliable ML workshops) and the arXiv preprint "Mind the Performance Gap: Capability-Behavior Trade-offs in Feature Steering."