Research
Research positions, original preprints and working papers, talks, and ongoing investigations.
1.Positions
Mar 2024 – Present
Undergraduate Researcher — Algorithms & Computational Geometry
GADA Lab (Geometric, Approximation & Distributed Algorithms) · University of Manitoba
Supervisors: Dr. Stephane Durocher, Dr. Anurag Murty Naredla, Dr. Leqi (Jimmy) Zhu & Dr. Avery Miller
- –Co-authoring Geometric Algorithms and Hardness for Fuel-Efficient Collaborative Delivery, which studies how cooperating delivery agents (e.g. drones) can move a package while minimizing total fuel use.
- –Results include an O(n²) algorithm for the line-restricted setting, improving the previous O(n³) bound from WALCOM 2025, and an O(n³) algorithm for the unrestricted L1 setting.
- –We prove the Euclidean (L2) setting is NP-hard, and show that the L1 algorithm gives a √2-approximation for it.
- –Developed an approximation algorithm for Hamiltonian-path optimization in spatially embedded graphs, with visualization and randomized testing across 100+ synthetic instances.
- –Improved the performance of a noncrossing spanning-tree research library by 30% and introduced stronger Python testing, documentation, and development practices.
- –Gave reading-group talks at weekly GADA lab meetings (see Talks).
Sep 2026 – Apr 2027
Honours Research Project — Calibrated Risk Prediction
Department of Computer Science · University of Manitoba
Supervisors: Dr. Christopher Henry, Dr. Mohammad Sayari & Dr. Sadaf Salehkalaibar
- –Building calibrated models that predict potato pathogen risk and yield loss, so that predicted probabilities can be trusted when making decisions under uncertainty.
- –Interdisciplinary project in collaboration with the Department of Plant Science.
May 2025 – Present
Machine Learning Researcher
Terrabyte Machine Learning Research Group · Winnipeg
- –Extended diffusion-based learning to higher-order relational data by implementing 4 Hypergraph Neural Diffusion variants in PyTorch and PyTorch Geometric.
- –Built a reproducible MLflow pipeline tracking configurations, metrics, and evaluations for 25+ experiments across 5+ datasets.
Jan 2024 – Jul 2025
Undergraduate Researcher — Combinatorics & Graph Theory
Manitoba eXperimental Mathematics Laboratory (MXML) · Department of Mathematics, University of Manitoba
Supervisors: Dr. Andrii Arman & Dr. Mahsa N. Shirazi
- –Established a coefficientwise inequality underlying the Hurlbert–Kamat property for perfect m-ary trees, and derived a closed-form expression for their independence number.
- –Used generating functions, polynomial recurrences, induction, and coefficientwise analysis to build the argument.
- –Ran SageMath and Python experiments across 1,000+ instances to test conjectures, verify symbolic identities, and validate intermediate results.
2.Preprints & Working Papers
Not yet peer-reviewed. Author order is alphabetical, as is standard in theoretical computer science.
- [1] S. Durocher, A. Maharjan, A. Miller, A. M. Naredla. Geometric Algorithms and Hardness for Fuel-Efficient Collaborative Delivery. Manuscript in preparation. Computational Geometry · Algorithms. An O(n²) algorithm for the line-restricted setting (improving the O(n³) bound of WALCOM 2025), an O(n³) algorithm for the L1 setting, NP-hardness for the Euclidean (L2) setting, and a √2-approximation for L2.
- [2] A. Arman, A. Maharjan, M. N. Shirazi. HK Property in Perfect Binary Trees. Preprint. Combinatorics · Graph Theory. A generating-function framework for counting the sizes of stars (independent sets containing a fixed vertex) in perfect m-ary trees. It gives an alternate, coefficientwise proof of the HK-property for perfect m-ary trees, first proved by Iľkovič and Yan with a combinatorial injection. It also measures the gap between the leaf star and the root star through a structured recurrence, and recovers EKR consequences such as formulas for the independence number and the minimum size of a maximal independent set. [PDF]
- [3] A. Maharjan. Wiener Index. Report. Graph Theory · Heuristics. Implementation of a heuristic algorithm for an open problem on the Wiener index (the sum of all pairwise distances between vertices of a graph), with an analysis of the problem. Exploratory; it does not resolve the problem. [PDF]
3.Talks & Presentations
Talks marked with an asterisk (*) present papers authored by others, given as reading-group talks.
Jul 28, 2025
Spectral Graph Theory for Maze Solving *
GADA Lab Research Meetings · EITC E2, University of Manitoba
Oct 3, 2025
Wiggly Permutations *
GADA Lab Research Meetings · University of Manitoba
Talk presenting the CCCG 2025 paper by its authors; presented as a reading-group talk.
Feb 4, 2026
Space-Efficient Eulerian Circuits *
GADA Lab Research Meetings · University of Manitoba
Talk presenting the SOSA 2026 paper by its authors; presented as a reading-group talk.
Mar 28, 2024
Exploring Perfect Binary Trees with Relation to the HK-Property
UM Combinatorics Seminar · Machray Hall, University of Manitoba
Sep 18, 2024
4.Ongoing Work
Fuel-Efficient Collaborative Delivery. Writing up the manuscript with Dr. Stephane Durocher, Dr. Anurag Murty Naredla & Dr. Avery Miller, and exploring an implementation and benchmark of the line-restricted algorithm.
Calibrated Pathogen-Risk Prediction. Honours project with Dr. Christopher Henry on calibrated prediction of potato pathogen risk and yield loss — learning and decision-making under uncertainty.
Hurlbert–Kamat Property for Trees. Continuing the investigation of the Hurlbert–Kamat property for tree families beyond the perfect binary case.
Hypergraph Neural Networks. Extending diffusion-based GNN architectures to hypergraphs with PyTorch Geometric.