PhD Dissertation Proposal by SHI Jieke | Optimizing and Fortifying AI Software via Program Synthesis

Optimizing and Fortifying AI Software via Program Synthesis

ABOUT THE TALK

Artificial Intelligence (AI) has fundamentally transformed modern software systems. Notable examples include the integration of Large Language Models (LLMs) into software engineering tools and the deployment of deep neural networks in autonomous vehicles. These advances define two major categories of AI software: AI4SE (AI for Software Engineering) and AI4Control (AI for Control Systems). While offering unprecedented capabilities, they also introduce an emerging AI software crisis, characterized by high resource consumption and unsafe behaviors, particularly in resource-constrained or mission-critical settings. 

This dissertation proposal addresses this crisis by advocating program synthesis, a technique for automatically generating programs that provably satisfy high-level specifications or optimize specific objectives, as a unifying methodology. Through five completed studies, we treat the configurations of AI models, auxiliary analyzers, and runtime monitors as artifacts to be synthesized, and demonstrate how program synthesis can serve as a versatile tool for (1) compressing and improving the efficiency of LLMs (AI4SE), and (2) synthesizing proxy programs and runtime shields to ensure the safe operation of control systems governed by neural controllers (AI4Control). Collectively, these contributions lay the groundwork for building intelligent systems that are efficient, reliable, and trustworthy by design.

SPEAKER BIOGRAPHY

SHI Jieke is a Ph.D. candidate and Research Engineer at the School of Computing and Information Systems (SCIS), Singapore Management University (SMU), under the supervision of Professor David LO, an ACM/IEEE/ASE Fellow. His research focuses on the intersection of Software Engineering (SE) and Artificial Intelligence (AI), with a particular focus on program synthesis to improve the efficiency and reliability of AI-powered software. His research has led to multiple publications at top-tier SE and security venues, including ICSE, ASE, TSE, TOSEM, and IEEE S&P. His work has been recognized with several academic honors, including an ACM SIGSOFT Distinguished Paper Award Nomination (ASE’22), an Honorable Mention Award (ACSAC’22), and an OpenAI Researcher Access Program Grant. At SMU, he has also received the SMU Presidential Doctoral Fellowship (2023), the SCIS Research Excellence Award (2025), and the SCIS Dean's List Award (2023 & 2025). More info: https://jiekeshi.notion.site