Publications
RuleFlow: Generating Reusable Program Optimizations with LLMs
AgentRx: Diagnosing AI Agent Failures from Execution Trajectories
Unified Operation Formalism for LLM-based Theorem Proving Systems
Lumos: Let there be Language Model System Certification
Cost-Driven Synthesis of Sound Abstract Interpreters
A Tensor-Based Compiler and a Runtime for Neuron-Level DNN Certifier Specifications
Efficient Ranking Function-Based Termination Analysis with Bi-Directional Feedback
Safety and Trust in Artificial Intelligence with Abstract Interpretation
Automated Verification of Soundness of DNN Certifiers
ConstraintFlow: A DSL for Specification and Verification of Neural Network Analyses
Interpreting Robustness Proofs of Deep Neural Networks