Mohammadreza Shakouri

Summary

I am a Computer Engineering graduate of Shahid Bahonar University of Kerman , and I am about to begin my master's degree at University of Tehran.

My research journey began during my bachelor's studies, where, under the guidance of Professor Mahdi Eftekhari, I worked on self-supervised learning for medical image analysis. This led to DINO-CXR, a vision-transformer-based self-supervised method for chest X-ray classification.

My interests center on multimodal learning with vision-language models, trustworthy AI, and AI for healthcare. I am curious by nature and enjoy exploring new research directions instead of staying within one fixed area. I am committed to applying what I learn to meaningful real-world problems, both through teamwork and independent work, and I look forward to engaging with international research groups and taking part in the wider research community.

Interests

Medical Image Analysis
Vision Language Models
Machine Learning
Computer Vision

Education

B.Sc.in Computer Engineering, 2018-2023

Shahid Bahonar University of Kerman

GPA: 3.23/4

Research Experience

Jan 2025 – Present

Researcher, Shahid Bahonar University of Kerman

Vision-Language Models for Treatment Planning in Restorative Dentistry

Supervised by Prof. Mahdi Eftekhari

Developing a multimodal vision-language pipeline for restorative dental treatment planning.
Investigating a clinical reasoning pipeline that extracts structured findings and uses an LLM for treatment planning.
Exploring retrieval-augmented generation (RAG) to ground responses in dental literature.

Sep 2022 – Aug 2023

Undergraduate Research Assistant, Shahid Bahonar University of Kerman

Self-supervised learning for chest X-ray classification

Supervised by Prof. Mahdi Eftekhari

Compared self-supervised methods with different networks and frameworks.
Designed and implemented a self-supervised method based on a vision transformer.

Publications

DINO-CXR: A Self Supervised Method Based on Vision Transformer for Chest X-Ray Classification

Springer, ISVC 2023

DINO-CXR adapts DINO, a self-supervised learning framework originally built for natural images, to a vision transformer backbone for chest X-ray classification. It tackles a real bottleneck in medical imaging — the scarcity of labeled data — by learning strong representations from unlabeled X-rays first. On pneumonia and COVID-19 detection, it outperforms existing self-supervised methods in accuracy while needing far less labeled data to fine-tune.

Link to paper

Teaching Experience

Fall 2021 - Spring 2022

Teaching Assistant — Advanced Programming

Department of Computer Engineering, Shahid Bahonar University of Kerman, Iran

Held weekly practice sessions for ~75 students with additional exercises to reinforce lecture material.
Designed and graded programming assignments and projects, providing feedback on code quality and design choices.

Fall 2022

Teaching Assistant — Artificial Intelligence and Expert Systems

Department of Computer Engineering, Shahid Bahonar University of Kerman, Iran

Graded assignments and exams, and gave feedback on projects.

Selected Projects

Principal Component Analysis (PCA) on the Mnist dataset

Performed PCA with SVD to determine the optimal number of principal components.

PyTorch Numpy Pandas Scipy

Decision Tree: ID3 Algorithm

Implemented decision trees using ID3 algorithm from scratch.

PyTorch Numpy Pandas

Medical MNIST Classification using VGG16/VGG19 and Transfer Learning

Conducted classification on the Medical MNIST dataset using transfer learning with VGG16/VGG19.

Python Numpy Pandas scikit-learn Matplotlib

Stochastic Gradient Descent (SGD) and Batch Gradient Descent algorithms for linear regression

Implemented linear regression using stochastic gradient descent and batch gradient descent.

Python Numpy Pandas Matplotlib

Neural Network Implementation for MultiClass Classification

Implemented neural network multiclass classification in Python from scratch.

Python Numpy Pandas scikit-learn Matplotlib

Frequency Domain Filtering

Implemented different filters (e.g., Gaussian) for frequency domain filtering.

Python OpenCV Numpy Matplotlib

US Airline Twitter Sentiment Analysis

Implemented an LSTM model to conduct sentiment analysis on US airlines Twitter database.

Python Numpy Pandas Nltk Seaborn

Recommender System

Implemented a recommender system based on DigiKala (An Iranian online shopping website)

Python gRPC