Gail McKinney

Hi, I’m

Gail McKinney

I'm a math and computer science student at Yale University. I work on AI assurance, an evolving field focused on rigorously evaluating the safety, capability, and behavior of AI systems.

I spent two internships at the Johns Hopkins University Applied Physics Laboratory building evaluation methods for operational and healthcare AI. The work I like best has to do with math, computer science, linguistics, and public health. My work is interdisciplinary because my grounding force is service of humanity. Humans are not confined to any single discipline.

Get to know me ↓ Email GitHub

Does a benchmark question actually tell you anything?

Discrimination 1.40 Difficulty 0.00 5 of 10 models clear it

This is an example item characteristic curve from a two-parameter logistic model. Each dot is an AI/ML model, and the curve gives its chance of getting one benchmark question right. When the curve is flat, every model scores about the same and the question can't tell them apart. A steep curve sorts the models. Working out what makes a benchmark or test strategy good for AI systems was a big part of my job at APL.

What I've built

Crash Cash

Personal project · 2026

A browser-based sandbox for learning personal finance through simulation. Take a job, open accounts, run up a credit card, file taxes, and watch what happens. It is designed to be a teaching tool for people who have limited access to financial education, and to be a safe space to make mistakes and learn from them.

  • JavaScript
  • Simulation
  • Unit testing
  • Agentic coding

Madeira Course Catalog

The Madeira School · 2026

A course planning app for students building a four-year academic plan. Browse the catalog, drag courses into grade year columns, and watch distribution and block requirements fill in as you go.

  • JavaScript
  • Interface design
  • Stakeholder work

Object Detection for Autonomous Driving

Inspirit AI Scholars · 2025

A five-model comparison for autonomous driving perception, presented as a review show so a mixed audience of students, parents, and instructors could follow the tradeoffs.

  • Python
  • PyTorch
  • Computer vision
  • Model evaluation

Learn more →

Get to know me

Education
Yale University The Madeira School
Experience
Johns Hopkins University Applied Physics Laboratory Riverside Baptist Church The Madeira School United States House of Representatives Fairfax County Public Schools
Projects
Crash Cash Madeira Course Catalog Object Detection for Autonomous Driving
Awards
Rising Star Award STEM Program Management Office Scholarship National Merit Finalist AP Scholar with Distinction Best Presentation Award

Elsewhere

My personal codebases (i.e. the projects linked on this site) are on GitHub. Much of my work is within the internal systems of organizations I've worked with. For more information, reach out via email or LinkedIn.