Reza Akbarian Bafghi

PhD Candidate, University of Colorado Boulder

I am a PhD candidate in Computer Science at the University of Colorado Boulder, advised by Prof. Maziar Raissi. My research focuses on scientific machine learning and computer vision, particularly physics-informed modeling, efficient model adaptation, and self-supervised learning, with applications to physical and biomedical systems. I also work on LLM alignment and agentic AI systems for scientific computing and robotics. My work has been published in TMLR, Medical Image Analysis, CVPR, WACV, NAACL, and ICCAS, as well as ICLR and NeurIPS workshops.

I have interned at Cruise (autonomous driving perception) and AngioInsight (medical AI for coronary hemodynamics), and collaborated with Ford (automotive engineering) and Vinci4D (physics AI).

selected publications

  1. centerlines.webp
    From Centerlines to Hemodynamics: Anisotropic RBF Decoders for Coronary Arteries
    Reza Akbarian Bafghi*, Sukirt Thakur*, and Maziar Raissi
    Transactions on Machine Learning Research, Aug 2026
  2. punch.webp
    PUNCH: Physics-informed Uncertainty-aware Network for Coronary Hemodynamics
    Sukirt Thakur, Marcus Roper, Yang Zhou, Dmitry Yu. Isaev, and 7 more authors
    Medical Image Analysis, Sep 2026
  3. aligning.webp
    Aligning to What? Limits to RLHF Based Alignment
    Logan Barnhart, Reza Akbarian Bafghi, Stephen Becker, and Maziar Raissi
    In Findings of the Association for Computational Linguistics: NAACL, 2025
  4. mixdiff.webp
    MixDiff: Mixing Natural and Synthetic Images for Robust Self-Supervised Representations
    Reza Akbarian Bafghi*, Nidhin Harilal*, Claire Monteleoni, and Maziar Raissi
    In IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2025
  5. peft.webp
    Parameter Efficient Fine-tuning of Self-supervised ViTs without Catastrophic Forgetting
    Reza Akbarian Bafghi*, Nidhin Harilal*, Claire Monteleoni, and Maziar Raissi
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshop, 2024
  6. vizwiz.webp
    A New Dataset Based on Images Taken by Blind People for Testing the Robustness of Image Classification Models Trained for ImageNet Categories
    Reza Akbarian Bafghi and Danna Gurari
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023

selected open-source projects

PINNs-Torch

Physics-informed neural networks implemented in PyTorch, with a focus on speed and usability.

Python PINNs-Torch GitHub stars

PINNs-TF2

A fast, user-friendly TensorFlow 2 implementation of physics-informed neural networks.

Python PINNs-TF2 GitHub stars

PINNs-JAX

A JAX implementation of physics-informed neural networks for composable scientific computing.

Python PINNs-JAX GitHub stars

PEFT-ViT

Parameter-efficient fine-tuning of self-supervised vision transformers without catastrophic forgetting.

Python PEFT-ViT GitHub stars

VizWiz Classification

A test set of images taken by blind people for evaluating ImageNet classification robustness.

Jupyter Notebook VizWiz Classification GitHub stars

Spotify Recommender

A music recommendation system built using the Spotify Million Playlist Dataset.

Python Spotify Recommender GitHub stars