Aryan Ahadinia

I am a Master’s student in Computer Science at EPFL and a Research Assistant at the ETH/EPFL AI Center (Swiss AI Initiative), where I work on inference and infrastructure for the Apertus LLM project.

I recently completed a six-month internship at Oracle, where I worked on AgentSpec and WayFlow, Oracle’s open-source libraries for building agentic workflows. Previously, I completed a semester research project at INDY at EPFL on Bayesian Optimization, which resulted in a publication at ICLR 2026. Before joining EPFL, I earned a Bachelor’s degree in Computer Engineering from Sharif University of Technology, where I completed my thesis on graph mining in social networks.

Research Interests

In a nutshell, I am interested in Machine Learning, with a particular focus on the following areas:

LLMs & Agents: During my internship at Oracle, I worked on various aspects of building, using, and evaluating LLMs and agents, with a particular focus on reasoning capabilities.
LLM Inference & Infrastructure: At the ETH/EPFL AI Center, I work on high-performance LLM serving and deployment infrastructure for the Apertus LLM project, running on the CSCS Alps supercomputing infrastructure.
Bayesian Optimization & Probabilistic Models: During a semester research project at EPFL, I worked on incorporating invariance into Bayesian Optimization to improve sample efficiency and achieve lower regret bounds, resulting in a publication at ICLR 2026. During my Bachelor’s, I worked on graph mining for social networks, with a focus on applying probabilistic models to time-series data to uncover hidden spatial and temporal relationships in social media, resulting in publications at arXiv (2023) and Scientific Reports (2024).

Education

M.Sc. in Computer Science, EPFL, 2024 - present.
B.Sc. in Computer Engineering, Sharif University of Technology, 2019 - 2023.

News

[July 2026] Swiss AI has released Apertus 1.5! Proud to be a part of this project and contribute to the development of this open-source LLM project!
[February 2026] Joining the ETH/EPFL AI Center (Swiss AI Initiative) as a research assistant, working on inference and infrastructure of the Apertus LLM project!
[January 2026] Our paper “Symmetry-Aware Bayesian Optimization via Max Kernels” is accepted at ICLR ‘26! This paper was the outcome of my semester project at INDY, EPFL, under supervision of Dr. Anthony Bardou and Prof. Dr. Patrick Thiran.
[January 2026] Happy to conclude my six month internship at Oracle!
[October 2025] Oracle has open-sourced agent-spec and wayflow for building agents! Proud to be a very small part of this project during my internship!
[July 2025] Starting a six month internship at Oracle! Excited to work on LLMs/agents there!
[September 2024] Starting master of Computer Science at EPFL!
[September 2023] Graduated from Department of Computer Engineering, Sharif University of Technology!

Publications

ICLR

Symmetry-Aware Bayesian Optimization via Max Kernels

Anthony Bardou, Antoine Gonon, Aryan Ahadinia, Patrick Thiran

The Fourteenth International Conference on Learning Representations (ICLR), 2026

PDF Project Page

Scientific Reports

DANI: Fast Diffusion Aware Network Inference with Preserving Topological Structure Property

Maryam Ramezani, Aryan Ahadinia, Erfan Farhadi, and Hamid R. Rabiee

Nature Scientific Reports volume 14, Article number: 31053 (2024)

PDF Code Project Page

arXiv

Joint Inference of Diffusion and Structure in Partially Observed Social Networks Using Coupled Matrix Factorization

Maryam Ramezani, Aryan Ahadinia, Amirmohammad Ziaei Bideh, and Hamid R. Rabiee

arXiv preprint arXiv:2010.01400, 2023

PDF Code Project Page