Team

We are researchers and red-teamers. Four of us are authors of Agents of Chaos (Shapira et al., 2026), the first study of emergent harm in populated multi-agent environments — covered by Science and Wired.

01/ network

02/ people

Alex Loftus

Lead

Founder and CEO

1st place winner of 1,249 teams, $1M Vesuvius Ink Detection (Kaggle)
Co-author, textbook on understanding networks (Cambridge UP)
Co-author of Agents of Chaos (Science, WIRED)
Co-Organized the New England mech-interp conference
Dropped interpretability Phd, David Bau lab

site scholar github

Jannik Brinkmann

Lead

Co-founder & Chief Scientist

Computer science PhD specializing in mechanistic interpretability and AI safety.
13+ research papers, with publications at NeurIPS, ICLR, and ICCV.
NAACL and EMNLP orals. EMNLP best-paper nominee. Honorable mention at the ICML 2024 Mechanistic Interpretability Workshop.
Research with David Bau’s lab (Northeastern), He He’s group (NYU), and J.P. Morgan AI Research.
Mech Interp Discord admin since 2023.

site scholar github

Antonio Mari

Lead

Co-founder & CTO

Leads technical delivery for our work with frontier AI labs.
Six years of machine learning research, specializing in reinforcement learning and AI-based cyberattack detection.
Recipient of EPFL’s Best GPA Award.
Published award-winning research at NeurIPS and ICML and TMLR.
ETH Zürich, EPFL, and Oracle.

scholar openreview

Tim Grams

Contractor

Reinforcement learning (ICLR 2025 spotlight) — University of Mannheim

frontier-lab campaigns · Mannheim

scholar

Robin Sieber

Contractor

ETH Zurich MSc, researching LLM safety preference alignment through reinforcement learning from human feedback. Published research on scalable dynamic programming optimization algorithms.
Built GPU-accelerated RL training pipelines and distributed computing systems using C++, Python, and CUDA.

linkedin github scholar

Alice Rigg

Contractor

Mechanistic interpretability researcher working with David Bau’s lab on NDIF. ICLR 2025 coauthor. Mech Interp Discord admin.

frontier-lab campaigns · EleutherAI · Bilinear MLPs

site scholar github

Negev Taglicht

Contractor

Interpretability researcher working with David Bau’s lab. Coauthor of Agents of Chaos.

Agents of Chaos · frontier-lab campaigns

linkedin dblp

Avery Yen

Contractor

Software engineer working on interpretability in David Bau’s lab. Coauthor of Agents of Chaos.

Agents of Chaos · frontier-lab campaigns · Northeastern