Smarter Code, Faster Software: Predicting Performance Bugs with AI
Presenter: Younes Jahandideh
Date: Thursday, November 26, 2026
Time: 2:30 pm – 4:30 pm
Location: Lecture Space/Workshop Space, 4th Floor - 300 Ouellette Ave., School of Computer Science Advanced Computing Hub
Abstract:
This interactive workshop introduces performance bugs and shows how static analysis, project history, and machine learning can help identify risky areas in software before they cause problems. Participants will explore how code and process features are extracted, how models such as Random Forest and XGBoost make predictions, and how to interpret those predictions. Through a group activity, they will assess software risk indicators and discuss the challenges of making reliable predictions. By the end, participants will be able to describe the prediction process and interpret its results while recognizing its limitations.
Workshop Outline:
Introduction and quick poll: What makes software slow?
Performance bugs and static analysis
Using project history to identify bug-prone code
Extracting code and process features
How Random Forest and XGBoost make predictions
Group activity: assess risk using sample software indicators
Interpreting results and discussing limitations
Wrap-up and questions
Prerequisites:
Basic familiarity with programming is helpful; no prior experience with performance bug prediction is required.
Biography:
Younes is a Ph.D. Candidate and research assistant who began his program at the School of Computer Science in Fall 2023. His primary research focuses on learning-based optimization, a type of self-optimizing system in cloud computing.