Katrina Brown

Hi! I'm Katrina, a pretraining researcher at Anthropic. I graduated magna cum laude from Harvard University ('25) with an undergraduate degree in Computer Science and Math, as well as a masters degree in Statistics. On campus, I conducted machine learning research, taught CS & Statistics courses, and helped run the undergraduate Women in Computer Science (WiCS) group.

▶︎ Outside of work

I grew up in Indiana, and since then I've passed my time living in San Francisco, Cambridge MA, NYC, and Chicago. In my spare moments, I love to hike the trails of SF, road bike, run half-marathons, ice skate, read, and infuse baked goods and smoothies with vegetables. My creative side manifests in violin, which I studied seriously under Mimi Zweig at the IU Jacobs School of Music from 5-16 years old, and more sporadically since.

My favorite pastime is meeting new people. Reach out to katr.rbrown [AT] gmail.com if you'd like to say hello!

GitHub/ LinkedIn/ X/ Substack

Katrina Ruth Brown

Research

I've been fortunate to be mentored by Finale Doshi-Velez at the Harvard Data to Actionable Knowledge (DtAK) Lab and Cynthia Dwork at the Harvard Kempner Institute.

* denotes equal contribution.

Set-Based Prompting: Provably Solving the Language Model Order Dependency Problem
Reid McIlroy and Katrina Brown et. al.
NeurIPS 2024, Vancouver, Canada
paper
Predictive Scheduling for Efficient Inference-Time Reasoning in Large Language Models
Katrina Brown*, Aneesh Muppidi*, and Rana Shahout
ES-FoMo III Workshop, ICML 2025, Vancouver, Canada
project page / paper
Order Independence With Finetuning
Katrina Brown and Reid McIlroy
Bi-Align Workshop, ICLR 2025, Singapore
paper
Evolutionary Prompt Optimization Discovers Emergent Multimodal Reasoning Strategies in Vision-Language Models
Sid Bharthulwar*, John Rho*, and Katrina Brown*
Reasoning and Planning Workshop, ICLR 2025, Singapore
paper
Diverse Concept Proposals for Concept Bottleneck Models
Katrina Brown, Marton Havasi, Finale Doshi-Velez
Human Machine Collaboration and Teaming Workshop, ICML 2022, Hawaii
paper

Writing and Media

Writing

Debate

In high school, I was a 2x state champion in Public Forum Debate. I was awarded the Bob Brittain Debate Scholarship (2021), which is granted to the single top debater in Indiana each year, and I was featured in the Herald Times as the first student from my school to win state.

Teaching

Teaching Courses

In Fall 2023, I was the Head Teaching Fellow for CS 121, our advanced theoretical computer science class with ~120 students. I've also TF'd Stat 110, our undergrad probability course, twice, as well as CS 120, our introductory algorithms course, and CS 124, our data structures and algorithms course, working with professors Boaz Barak, Joe Blitzstein, Madhu Sudan, and Adam Hesterberg.

Academics

Coursework

Harvard University: Systems Programming, Data Structures & Algorithms, Database Systems, Machine Learning, Reinforcement Learning, Sequential Decision Making, Foundation Models, Theoretical Computer Science, Algorithms at the End of the Wire, Distributed Systems, Probability Theory, Statistical Inference, Linear Models, Time Series, Generalized Linear Models, Linear Algebra, Multivariable Calculus, Real Analysis, Abstract Algebra, Topology

Usual Gamut of Academic Accolades

  • Harvard University GPA: 3.97/4.0
  • Graduated magna cum laude
  • Phi Beta Kappa - Inducted to the oldest academic honor society in the United States (top 10% of Harvard graduating class based on coursework rigor and merit)
  • John Harvard Scholar - Awarded to students in the top 5% of their class in terms of GPA.
  • Detur Book Prize Recipient - Awarded for outstanding academic performance during the first three semesters at Harvard College.
  • IAG Scholarship recipient - awarded annually to the top 2 students academically and extracurricularly in Indiana
  • NCWIT Regional Award recipient
  • National AP Scholar (14 AP Courses)
  • SAT: 1580/1600