Paul Krogmeier

I’m a postdoctoral fellow at Harvard advised by Nada Amin and Walter Fontana. My primary area of research touches programming languages, logic, and formal methods.

I will be joining the University of Colorado Boulder as an assistant professor in computer science starting Fall 2026. I received a PhD from the University of Illinois advised by Madhusudan Parthasarathy.

I’m also a jazz alto saxophone player 🎷 and always interested in chances to jam. Shoot me an email if you’d like to talk about anything below.

Current Interests

  • How to make machine learning models more interpretable
  • Computational universality with errors:

    How can processes with limited accuracy perform universal computations?

  • Formal expressivity and algorithmic capabilities of neural architecture classes
  • Evolution of language and connections between evolution and learning
  • Teaching computer science and music improvisation together:

    Learning to improvise within classes of "regular" riffing patterns, adding a "stack", etc.

  • Rigorous characterizations of creativity:

    What's the difference between creative processes and automatic processes?

Ongoing Projects

  • Emergence of symbolic language: developing a computational model in which symbolic language and abstraction are emergent outcomes of computation
  • DSL synthesis: theory and algorithms for automatically synthesizing DSLs that express relevant domain concepts succinctly and irrelevant ones less succinctly or not at all

Publications

Synthesizing DSLs for Few-shot Learning
Paul Krogmeier and P. Madhusudan
Conditionally accepted at OOPSLA 2025
Languages with Decidable Learning: A Meta-theorem
Paul Krogmeier and P. Madhusudan
★Distinguished Paper Award at OOPSLA 2023
Synthesizing Axiomatizations using Logic Learning
Paul Krogmeier*, Zhengyao Lin*, Adithya Murali*, and P. Madhusudan
OOPSLA 2022
Composing Neural Learning and Symbolic Reasoning with an Application to Visual Discrimination
Adithya Murali, Atharva Sehgal, Paul Krogmeier, and P. Madhusudan
IJCAI 2022
Learning Formulas in Finite Variable Logics
Paul Krogmeier and P. Madhusudan
★Distinguished Paper Award at POPL 2022
Deciding Accuracy of Differential Privacy Schemes
Gilles Barthe, Rohit Chadha, Paul Krogmeier, A. Prasad Sistla, and Mahesh Viswanathan
POPL 2021
Decidable Synthesis of Programs with Uninterpreted Functions
Paul Krogmeier, Umang Mathur, Adithya Murali, P. Madhusudan, and Mahesh Viswanathan
CAV 2020
Deciding Memory Safety for Single-Pass Heap-Manipulating Programs
Umang Mathur, Adithya Murali, Paul Krogmeier, P. Madhusudan, and Mahesh Viswanathan
POPL 2020
Towards Context-Aware Data Refinement
Paul Krogmeier, Steven Kidd, and Benjamin Delaware
Intl. Workshop on Coq for Programming Languages 2018

*equal contribution