Title Software Engineering Foundations for Reliable AI Systems

Abstract

AI systems built on deep neural networks now underpin critical applications in transportation, healthcare, security, and beyond. Yet unlike traditional software, their behaviors arise from data-driven statistical processes, making them difficult to test, interpret, and assure with classical engineering techniques. This talk presents a software engineering perspective on AI reliability: treating learning-enabled components as software artifacts whose behavior must be systematically exercised, perturbed, and analyzed. I will summarize recent advances in test adequacy for neural models, scalable mutation-guided testing, fault analysis, and system-level reliability, and discuss new challenges posed by foundation models such as LLMs, RAG-based LLM applications, and embodied multimodal agents.

Bio

Jinhan Kim is a postdoc at TAU lab in Università della Svizzera italiana (USI) led by Prof. Paolo Tonella. He completed his Ph.D. degree from KAIST under the supervision of Prof. Shin Yoo. His research bridges Software Engineering (SE) and Artificial Intelligence (AI), focusing on testing and reliability of AI systems. He develops principled methods to assess and improve the robustness of complex systems, from traditional software to LLMs, aiming to make them reliable and transparent for deployment in safety-critical domains. More information is available at https://jinhan.me/.