From data to robust surgery schedules: learning the uncertainty of surgical duration and recovery-room stay

Project

Master 2 internship within the ANR SMART-SURGE project, hosted by IBISC, Université Évry Paris-Saclay.

Location and schedule

• Location: IBGBI building, 23 boulevard de France, 91000 Évry-Courcouronnes, France
• Duration: 6 months, from February 2027 to July 2027
• Application deadline: November 15, 2026
• Supervisors: Salma Makboul, Alexandru-Liviu Olteanu, and Marc Sevaux

Context

Operating rooms are among the most expensive and constrained resources in hospitals. Elective surgery scheduling is difficult because planned schedules are frequently disrupted: operations may run late, patients may stay longer than expected in the post-anesthesia care unit (PACU), and staff may be unavailable. These events are interdependent, as operating-room delays can fill the PACU and hold up subsequent patients.

SMART-SURGE develops robust and distributionally robust optimization models for elective surgery scheduling, combined with machine learning, and validates them with a large public hospital in Essonne, France. The quality of these models depends on accurately describing uncertainty. A schedule based on poorly estimated durations may be too cautious, leaving operating rooms empty, or insufficiently cautious, resulting in overtime and cancellations.

Work plan

The intern will:

• Build a clean dataset from open perioperative sources and pseudonymized hospital data.
• Conduct exploratory analysis and address missing values and outliers.
• Encode clinical variables.
• Convert the main data tables into standard HL7 FHIR resources as a first step toward a reusable data pipeline.
• Develop predictive models for operation duration and PACU length of stay, ranging from simple baselines to machine-learning models.
• Evaluate the models under realistic conditions.
• Quantify uncertainty beyond point predictions by studying how actual durations deviate from expected durations and how this uncertainty varies across surgical contexts.
• Integrate the learned uncertainty into a surgery scheduling model under uncertainty.
• Assess the impact on schedule quality, including overtime, idle time, and PACU occupancy, compared with more classical assumptions.

The internship will conclude with a report and documentation of the data, models, and results, which will support the PhD student working on the project.

Profile

• M2 student in operations research, data science, computer science, or applied mathematics.
• Good command of Julia and/or Python.
• Some experience with an optimization solver such as Gurobi or CPLEX is appreciated.
• Solid foundations in statistics, machine learning, and linear/integer programming.
• Interest in healthcare applications.

Prior experience with medical data is not required.

Environment and benefits

• Combination of operations research and machine learning applied to a concrete healthcare problem.
• Dedicated high-performance workstation.
• Opportunity to present results at national and international conferences.
• Integration into the SMART-SURGE team as a project member.
• Natural route toward the PhD funded by the project at Université Paris-Saclay.