Mathematics & Computer Science
Amherst College
Efficient Neural Network Verification @ Amherst College
Developing Luna, a C++ bound propagator for efficient neural network verification. Implements Interval Bound Propagation, DeepPoly/CROWN, and α-CROWN over general computational graphs, outperforming the state-of-the-art Python implementation on VNN-COMP 2025 benchmarks in both bound tightness and runtime.
Awarded the Computer Science Prize at Amherst College, recognizing the most outstanding honors thesis and contribution to the department.
📄 NEAT: The Luna Bound Propagator for Formal Analysis of Neural Networks — under review at SAS 2026 (with Dr. Haoze Wu).