Nicole Ma

Hi! I'm an undergrad at Stanford studying CS on the AI track. I'm generally interested in AI alignment, mechanistic interpretability, and reasoning. I've worked on activation monitoring as a research fellow @ SPAR, reasoning and distillation as an undergrad researcher @ SAIL, spatial reasoning and physical optimization as co-founder @ Topological, and geophysical modeling as an intern @ NASA.

On the side, I like to write/edit/translate Wikipedia articles. In the past 3 years, articles I've worked on in my areas of interest (music, literature, technology, China, and the Middle East) have received 100M+ views. I'm involved in post-colonial and arabization research @ CCSRE.

Email  /  Github  /  LinkedIn  /  X

profile photo

Selected Projects

Optimizing Remasking Schedules for Reasoning in Discrete Diffusion Models

Diffusion LLM optimized for reasoning and efficiency at inference time with SFT + RL for dynamic diffusion step adaptation and remasking scheduling

SHIELD.

Multi-agent RL + tool use framework for detecting and patching security vulnerabilities in codebases

NeuroPilot

Brain-computer-interface and agentic AI system for brain-powered natural language commands for hands-free computer control

Synthetic Control Analysis of the Trade Effects of Diplomatic Recognition

Counterfactual analysis of dollar diplomacy using synthetic controls and causal inference

Selected Awards

1st Place ($40k) @ Mercor x Cognition x Etched Hackathon

Most Impressive Technical Feat @ Pear VC x Anthropic hackathon

1st Place Scrapybara Award ($16k) @ Treehacks

Various hackathon awards: 1st Place @ Google igniteCS Programming Expo, 1st Place @ GameGala, Best Web App @ HackTable, Best AR/VR App @ LightHouse Hacks, HM @ HackVH, One Percenter Award @ Project Euler

Various science fair awards: Finalist in Math/CS @ NJSHS, 1st Place @ JSHS, 3rd Place @ CSEF, 1st Place @ OCSEF, Association of Women in Water, Energy, and Environment Award, NASA Earth System Science Award, the Energy Coalition First Place Award

Design and source code from Jon Barron's website