Andy Wang

AI Safety Researcher | UW Madison CS & Math

Astra Research Fellow working on CoT monitorability with Scott Emmons (Anthropic, prev. Google DeepMind). Previously Research Fellow at SPAR and Research Intern at American Family Insurance. Also known as yeedrag online.

I think making sure AI goes well is probably the most important problem right now, and I'd like to spend my career helping with that.

Email · LinkedIn · GitHub · Discord: yeedrag

Experience

Astra Research Fellow

Jan 2026 – Present

Constellation (Berkeley, CA)

  • Working with Scott Emmons (Anthropic, prev. Google DeepMind) on CoT monitorability.

AI Safety Research Fellow

Sep 2025 – Dec 2025

Supervised Program for Alignment Research (Berkeley, CA)

  • Investigated various properties on CoT Monitorability, mentored by Qiyao Wei.

AI Security Research Intern

May 2025 – Dec 2025

American Family Insurance (Madison, WI)

  • Conducted red-team assessments on internal AI infrastructure, received return offer to develop model safeguards.

Deputy Director

May 2025 – Present

Wisconsin AI Safety Initiative

  • Reframing strategy to focus on producing AI safety professionals.
  • Connecting with researchers to refine strategies for safety-focused groups.

Interests

Research Areas

  • CoT Monitorability
  • AI Control
  • Eval Awareness
  • Misalignment & Generalization

Personal

  • Combinatorics
  • Number Theory
  • Video Games
  • Cooking