I lead the Constraint Solving ("ConSol") research group, where we develop combinatorial optimisation algorithms that can be applied to a wide range of (real-world) problems. I am also the co-director of the explainable AI in transportation lab ("XAIT") as part of the Delft AI Labs. Prior to my appointment at TU Delft, I worked as postdoc at the University of Melbourne (2017-2020), received my PhD from the Vienna University of Technology in 2017, and in between these positions held shorter term posts at a production planning and scheduling company MCP (Vienna, Austria) and the National Institute of Informatics (Toyko, Japan).
My focus now is on developing solving techniques based on constraint programming, optimising decision trees, and explainable methods for combinatorial optimisation. I am also interested in industrial applications, robust/resilient optimisation, and the integration of optimisation and machine learning.
My work is regularly published at leading AI conferences (e.g., AAAI, NeurIPS) and specialised venues (e.g., CP, CPAIOR). I have been invited to scientific events such as Dagstuhl seminars, Lorentz workshops, and the Simons-Berkeley programme. I enjoy organising both incoming and outgoing research visits (e.g., EPFL, ANITI/CNRS, CUHK, Monash University, TU Wien), and in general highly value collaboration with (inter)national researchers. In the past I participated in several algorithmic competitions, scoring first place, e.g., incomplete tracks of MaxSAT Evaluation 2018+, ROADEF/EURO 2012. Competitions serve as a good way of promoting my work beyond publications, e.g., Google OR-Tools adopted some of my ideas. Our approaches on optimising decision trees offer order-of-magnitude runtime improvements whilst support additional constraints. I am also happy to work on real-world problems together with academic and industrial partners.
My long term vision is to automate decision making that is currently done by humans, referring to scenarios where difficult decisions need to be made under complex constraints. This not only results in more efficient decision making and resource utilisation, but also relieves human experts of the burden of handling difficult and stressful tasks, leaving them more time to deal with creative and meaningful problems that have not (yet) been automatised.
Office:
4.E.400, Building 28
Email: e.demirovic@tudelft.nl
Publications: Google Scholar
Research Software Engineer
PhD candidate, optimal decision trees
PhD candidate, constraint programming
PhD candidate, ride-sharing, forecasting
PhD candidate, explainable
combinatorial optimisation
PhD candidate, scheduling for quantum
Postdoc (2022), scheduling
Core-guided search for constraint programming
Column generation for employee scheduling
Conflict analysis for constraint programming
MaxSAT
Train shunting with constraint programming
MaxSAT, now PhD student
Constraint Programming
Scheduling for Quantum Computers
Traffic Predictions
SAT for Multi-Agent Path Finding
Explainable Predictive Maintainance
Automating Puzzle Generation
Graph Theory
(BSc honours) Algorithm Selection
Multi-Agent Path Finding
Reinforcement Learning for Logistics
MaxSAT for Correlation Clustering
MaxSAT
Predictions for Intensive Care
Algorithm Selection