Reza Mirjalili

Houston, Texas

Open to research and industry roles

I build algorithms that make large-scale logistics decisions tractable — and, increasingly, algorithms that learn how to make themselves faster.

My doctoral work at the Systems Optimization and Computing Laboratory, under Prof. Gino J. Lim, produced exact methods for drone-assisted delivery: branch-and-price-and-cut frameworks whose pricing subproblems are solved by bidirectional labeling, and whose search is guided by reinforcement learning rather than by hand-tuned rules. The aim throughout is provable optimality on instances that were previously out of reach.

The problem in one minute — a truck-only tour against a coordinated mothership-and-drone schedule, and the column-generation loop that finds it.

what I work on

Exact methods at scale

Branch-and-price-and-cut for path-based routing formulations. Column generation where the pricing subproblem is an elementary shortest path problem with resource constraints, solved by bidirectional label setting.

Cutting planes

Chvátal–Gomory, cover, conflict, and cycle-elimination inequalities — including a hybrid family that closed 8% more of the optimality gap than any of its components alone.

Learning inside the solver

Deep reinforcement learning for pricing, branching, and cut selection. Graph neural networks that prioritize subproblems, cutting column-generation convergence time by 20–30%.

Logistics applications

Drone-assisted and multi-echelon delivery networks, healthcare scheduling, and transportation network resilience under disruption.

Browse projects Publications Full CV

toolbox

Optimization — Gurobi, CPLEX, SCIP, Pyomo, column generation, branch-and-price · AI/ML — PyTorch, TensorFlow, deep RL, graph neural networks, CUDA · Languages — Python, C, C++, Rust, Cython, MATLAB · Parallelism — OpenMP, MPI, CUDA, Cython nogil

news

Nov 19, 2025	New preprint on arXiv: Resource-Based Time and Cost Prediction in Project Networks, applying graph neural networks to project scheduling.
Aug 15, 2025	Defended and completed my Ph.D. in Industrial Engineering at the University of Houston.
Jun 01, 2025	Completed the MLOps and LLMOps specializations at Duke University.
Nov 10, 2024	Earned the NVIDIA certifications in CUDA C/C++ Programming and Deep Learning Fundamentals.

selected publications

Working paper

Coordinated Mothership Scheduling and Drone Routing Optimization: A Branch-and-Price-and-Cut Approach

Reza Mirjalili and Gino J. Lim

2026

Manuscript in preparation

Bib

@unpublished{mirjalili2026mothership,
  title = {Coordinated Mothership Scheduling and Drone Routing Optimization: A Branch-and-Price-and-Cut Approach},
  author = {Mirjalili, Reza and Lim, Gino J.},
  year = {2026},
  note = {Manuscript in preparation},
}

arXiv
Resource-Based Time and Cost Prediction in Project Networks: From Statistical Modeling to Graph Neural Networks

Reza Mirjalili, Behrad Braghi, and Shahram Shadrokh Sikari

arXiv preprint arXiv:2511.15003, Nov 2025

Abs DOI arXiv Bib HTML PDF

Traditional project-planning methods such as CPM and PERT rely on simplified and often static assumptions regarding task interdependencies and resource performance. This work models projects as heterogeneous activity–resource graphs and evaluates GraphSAGE and Temporal Graph Networks for predicting project duration and cost. The learned models achieve a 23 to 31 percent reduction in mean absolute error relative to conventional baselines, with R² improving to approximately 0.91 on complex projects, and the learned embeddings provide interpretable information about resource constraints and critical dependencies.
@article{mirjalili2025resource, title = {Resource-Based Time and Cost Prediction in Project Networks: From Statistical Modeling to Graph Neural Networks}, author = {Mirjalili, Reza and Braghi, Behrad and Shadrokh Sikari, Shahram}, journal = {arXiv preprint arXiv:2511.15003}, year = {2025}, month = nov, doi = {10.48550/arXiv.2511.15003}, url = {https://arxiv.org/abs/2511.15003}, }
TRR
Resilience Analysis of New York City Transportation Network After Snow Storms

Reza Mirjalili, Hojjat Barati, and Anil Yazici

Transportation Research Record: Journal of the Transportation Research Board, Nov 2023

Abs DOI Bib HTML

This study examines how New York City’s transportation network recovers after snow storms, using network science metrics together with public data from the city’s sanitation department and traffic speed detectors across eight snow events. Bhattacharyya distance and Kolmogorov–Smirnov tests are used to determine when traffic conditions return to normal patterns. The results show that less data-intensive graph theory metrics can be used to estimate transportation network performance, enabling recovery-time forecasts that help city officials and businesses anticipate when operations can resume after winter storms.
@article{mirjalili2023resilience, title = {Resilience Analysis of New York City Transportation Network After Snow Storms}, author = {Mirjalili, Reza and Barati, Hojjat and Yazici, Anil}, journal = {Transportation Research Record: Journal of the Transportation Research Board}, volume = {2677}, number = {1}, pages = {694--707}, year = {2023}, publisher = {SAGE Publications}, doi = {10.1177/03611981221101034}, url = {https://doi.org/10.1177/03611981221101034}, }
JINT
Scheduling Diagnostic Testing Kit Deliveries with the Mothership and Drone Routing Problem

Hyung Jin Park, Reza Mirjalili, Murray J. Côté, and 1 more author

Journal of Intelligent & Robotic Systems, Nov 2022

Abs DOI Bib HTML PDF

A critical component in the public health response to pandemics is the ability to determine the spread of diseases via diagnostic testing kits. This paper presents a delivery scheduling method for diagnostic testing kits using a truck that carries multiple drones. The problem is decomposed into two phases: optimizing the truck route to minimize travel distance, then optimizing the drone delivery schedule to minimize total delivery time, iterating until an optimal solution is reached. Heuristic algorithms reduce computation time to under 50 minutes compared with over 10 hours for exact methods, and against conventional drive-through testing sites the drone-based approach achieves a modified basic reproduction rate of 0.002 versus 0.0153 for face-to-face testing.
@article{park2022scheduling, title = {Scheduling Diagnostic Testing Kit Deliveries with the Mothership and Drone Routing Problem}, author = {Park, Hyung Jin and Mirjalili, Reza and C{\o}t{\'e}, Murray J. and Lim, Gino J.}, journal = {Journal of Intelligent \& Robotic Systems}, volume = {105}, number = {2}, pages = {38}, year = {2022}, publisher = {Springer}, doi = {10.1007/s10846-022-01632-1}, url = {https://doi.org/10.1007/s10846-022-01632-1}, }