Practical DRAMSys for AI Inference (1st Offering) - JLR Challenge #2 Technical Workshop By: Sadra Hakim

 

Practical DRAMSys for AI Inference (1^st Offering)

Presenter: Sadra Hakim

 

Date: Wednesday, October 29, 2025

Time: 10:00 am

Location: Workshop Space, 4th Floor - 300 Ouellette Ave., School of Computer Science, Advanced Computing Hub

 

This workshop session begins with a tour of DRAMSys’s internal concepts and flow, and how to interpret its output metrics. In the second part, we connect those ideas to AI inference, where we profile a standard AI model on CPU and GPU, translate observed behaviour into DRAMSys scenarios, and compare a few representative configurations. By the end, participants will be able to read DRAMSys metrics with confidence and apply them to justify a memory configuration that improves AI inference performance.
 

1. Brief tour of internal concepts and components of DRAMSys.
2. Profile an AI run and translate observed behaviour into DRAMSys parameters.
3. Compare a few configs, interpret results, and outline how to justify the optimal setup for inference.

 

Familiarity with Python and Linux commands would help, but is not required.

 

Sadra is a Ph.D. student in Computer Science at the University of Windsor, where his research focuses on applying machine learning and deep learning models to biomedical challenges.

 

Registration Link (Only MAC students need to pre-register)