Reza Arabpour

Faculty Affiliate Researcher

Vector Institute for AI

Toronto, ON, Canada

Email: arabpour@mcmaster.ca

Engaging Summary

I’m a Computational Science & Engineering researcher passionate about making artificial intelligence more accessible and effective. Currently pursuing my M.Sc. at McMaster University with a 3.9/4.0 GPA, I focus on the exciting intersection of geometric deep learning and computational finance.

Current Work

Research Focus

As a Faculty Affiliate Researcher at Vector Institute, I develop innovative approaches to machine learning that tackle real-world challenges:

Parameter-Efficient Fine-Tuning (PEFT): Creating methods to make large language models more accessible and sustainable
Geometric Deep Learning: Building hypernetworks that understand complex financial processes
High-Performance Computing: Optimizing algorithms for multi-GPU environments

Academic Role

At McMaster University, I serve as a Graduate Teaching Assistant for advanced courses in:

Risk Management (MFM 714)
Computational Finance (MFM 713)
Numerical Methods in Finance (MFin 704)

Supporting over 500 undergraduate students in Python programming and organizing bi-weekly CSE seminars that bridge cutting-edge theory with practical applications.

Background & Achievements

Education & Recognition

M.Sc. Computational Science & Engineering (McMaster University, 2023-2025)
- Full scholarship recipient with 3.9/4.0 GPA
B.Sc. Applied Mathematics (University of Tehran, 2018-2023)
- Bronze medal in ACM ICPC West Asia Regional Contest

Entrepreneurial Experience

Co-founded a fintech startup where I engineered evolutionary-algorithm-driven trading strategies and conducted extensive Monte Carlo backtesting to optimize risk-adjusted returns across multiple market conditions.

Recent Highlights

ICML 2025 Workshop: Paper accepted on CPU-efficient LLM fine-tuning
Analysis & Applications: Published research on Volterra processes
Vector Institute: Faculty Affiliate Researcher position
Academic Excellence: Maintained top performance throughout graduate studies

Research Philosophy

I believe in making AI more accessible and sustainable while solving complex real-world problems. My work bridges theoretical advances with practical applications, particularly in financial modeling and efficient machine learning systems.

Always keen to connect with fellow researchers and practitioners in deep learning and finance.

news

Jul 02, 2025	Our paper “LoRA Fine-Tuning Without GPUs” has been accepted to the ICML 2025 Workshop on Efficient Systems for Foundation Models! 🎉
Jun 30, 2025	Our paper “LoRA Fine-Tuning Without GPUs: A CPU-Efficient Meta-Generation Framework for LLMs” is now available on arXiv! 📄
Jul 15, 2024	Completed the Deep Learning + Reinforcement Learning Summer School at CIFAR! Great opportunity to learn cutting-edge techniques and network with fellow researchers. 🧠
May 30, 2024	Our research on “Low-dimensional approximations of the conditional law of Volterra processes” is now available on arXiv and under review! 📝
Mar 15, 2024	Presented research on geometric deep learning applications in finance at the Fields Institute seminar series! 🎯

selected publications

Under Review
Low-dimensional approximations of the conditional law of Volterra processes: a non-positive curvature approach

Reza Arabpour, John Armstrong, Luca Galimberti, and 2 more authors

Under review at Analysis and Applications, May 2024

Abs Bib HTML

Predicting the conditional evolution of Volterra processes with stochastic volatility is a crucial challenge in mathematical finance. While deep neural network models offer promise in approximating the conditional law of such processes, their effectiveness is hindered by the curse of dimensionality caused by the infinite dimensionality and non-smooth nature of these problems. To address this, we propose a two-step solution. Firstly, we develop a stable dimension reduction technique, projecting the law of a reasonably broad class of Volterra process onto a low-dimensional statistical manifold of non-positive sectional curvature. Next, we introduce a sequentially deep learning model tailored to the manifold’s geometry, which we show can approximate the projected conditional law of the Volterra process. Our model leverages an auxiliary hypernetwork to dynamically update its internal parameters, allowing it to encode non-stationary dynamics of the Volterra process, and it can be interpreted as a gating mechanism in a mixture of expert models where each expert is specialized at a specific point in time.
@article{arabpour2024volterra, title = {Low-dimensional approximations of the conditional law of Volterra processes: a non-positive curvature approach}, author = {Arabpour, Reza and Armstrong, John and Galimberti, Luca and Kratsios, Anastasis and Livieri, Giulia}, journal = {Under review at Analysis and Applications}, year = {2024}, month = may, eprint = {2405.20094}, archiveprefix = {arXiv}, primaryclass = {math.NA}, url = {https://arxiv.org/abs/2405.20094}, }
ICML 2025
LoRA Fine-Tuning Without GPUs: A CPU-Efficient Meta-Generation Framework for LLMs

Reza Arabpour, Haitz Ocáriz Borde, and Anastasis Kratsios

ICML 2025 Workshop on Efficient Systems for Foundation Models, Jul 2025

Awarded Abs Bib HTML

Accepted to ICML 2025 Workshop on Efficient Systems for Foundation Models

Low-Rank Adapters (LoRAs) have transformed the fine-tuning of Large Language Models (LLMs) by enabling parameter-efficient updates. However, their widespread adoption remains limited by the reliance on GPU-based training. In this work, we propose a theoretically grounded approach to LoRA fine-tuning designed specifically for users with limited computational resources, particularly those restricted to standard laptop CPUs. Our method learns a meta-operator that maps any input dataset, represented as a probability distribution, to a set of LoRA weights by leveraging a large bank of pre-trained adapters for the Mistral-7B-Instruct-v0.2 model. Instead of performing new gradient-based updates, our pipeline constructs adapters via lightweight combinations of existing LoRAs directly on CPU.
@article{arabpour2025lora, title = {LoRA Fine-Tuning Without GPUs: A CPU-Efficient Meta-Generation Framework for LLMs}, author = {Arabpour, Reza and Sáez de Ocáriz Borde, Haitz and Kratsios, Anastasis}, journal = {ICML 2025 Workshop on Efficient Systems for Foundation Models}, year = {2025}, month = jul, eprint = {2507.01806}, archiveprefix = {arXiv}, primaryclass = {cs.LG}, url = {https://arxiv.org/abs/2507.01806}, }