AI4DA Speaker Series

7 September 2026

From Elegant Models to Messy Decisions: Taking Analytics from the Academic Problem to the Real World

Roozbeh Yousefi is a researcher and Senior Decision Scientist at Ontario Health specializing in analytics, operations research, forecasting, simulation, and data-driven decision-making. He holds a PhD in Analytics from Smith School of Business (Queen’s University) and has extensive experience developing quantitative models for complex, real-world systems.

His research interests include time-series forecasting, queuing systems, optimization, and partially observable Markov decision processes (POMDPs), with applications in healthcare systems, environmental management, and conservation of endangered species. His current research focuses on healthcare systems and population-level demand, particularly modeling Alternate Level of Care (ALC), healthcare capacity, patient flow, and the effects of system-level interventions.

More broadly, his work seeks to integrate advanced analytical and computational methods to support decision-making under uncertainty and translate quantitative models into practical decision-support frameworks.

 

Abstract

Academic training in Analytics teaches us to formulate precise problems, develop rigorous models, and evaluate solutions under clearly defined assumptions. Real-world decision problems rarely offer the same luxury. They emerge from complex systems, incomplete information, organizational constraints, competing objectives, and questions that are often difficult to formulate in mathematical terms.

In this talk, I will reflect on the transition from solving academic problems to addressing real-world decision problems, drawing on my experiences across operations research, forecasting, simulation, healthcare analytics, and decision-making under uncertainty. I will discuss how ideas such as partially observable Markov decision processes (POMDPs), optimization, and simulation can move from theoretical research into practical applications—including conservation of endangered species and healthcare system planning.

Using healthcare demand forecasting and Alternate Level of Care (ALC) modeling as a central example, I will illustrate how a seemingly straightforward research question, i.e., “How many patients will we need to accommodate?” quickly becomes a much broader decision problem involving population dynamics, patient flow, discharge capacity, behavioral and social factors, uncertainty, and policy interventions.

The challenge is not simply to build a more sophisticated model, but to determine what should be modeled, what can be measured, what assumptions are defensible, and how model outputs can support decisions.

Event Details

Date: September 18, 2026
Time: 3:30–4:30 PM MDT
Location: Virtual
RSVP deadline: September 15

RSVP