david minh-duy cao

Principles of Programming Languages (POPL) 2023

babble: Learning Better Abstractions with E-Graphs and Anti-Unification

David Cao*, Rose Kunkel*, Chandrakana Nandi, Max Willsey, Zachary Tatlock, Nadia Polikarpova

Humans are really good at library learning: finding common patterns in code and extracting them into a library of common abstractions. This problem has a bunch of use cases across FPGA design, program synthesis, and modeling human perception. So how can we get a computer to do this?

Building off of prior work in library learning and e-graphs, we contribute new algorithms for proposing candidate abstractions and choosing the best set. We improve on the prior state-of-the-art in compression and speed by several orders of magnitude!

Paper Slides (.key) Code (Github)

Programming Language Design and Implementation (PLDI) 2022

Type-Directed Program Synthesis for RESTful APIs

Zheng Guo, David Cao, Davin Tjong, Jean Yang, Cole Schlesinger, Nadia Polikarpova

Interacting with REST APIs can often be extremely unwieldy; APIs can have hundreds of methods, and may or may not be documented. We propose a system that can synthesize programs that call into REST APIs.

Paper

POPL 2021 • Student Research Competition, 2nd Prize

Automated Dependent Resource Analysis

David Cao, advised by Tristan Knoth and Nadia Polikarpova

When programming, it's useful to know the resource use of a program (in undergrad we take whole courses on this!). We contribute a system that can automatically infer resource bounds that can depend on the inputs of a program.

Abstract Virtual poster (video)