LinkedIn ↗
Google Scholar ↗
ORCID ↗01
Alzheimer's Research
Using large language models and artificial intelligence to support early detection, biomarker discovery, and therapeutic research in Alzheimer's disease.
LLMsHealthcare AIDrug Discovery
Biomedical AI Researcher
Shahidur Rahoman Sohag
I study Alzheimer's disease with biomedical AI, connecting large language models, multimodal learning, and trustworthy reasoning for diagnosis and therapeutic discovery.
LLM
CAUSAL AI
SECURE
ALZHEIMER'S
FAIRNESS AI
AIBIOMEDICAL
MISSIONDiagnosis
DISCOVERYDrug Repurposing
IMPACTPrecision Medicine
Reasoning for better healthcare
7Peer-reviewed publications
120+Students mentored
20%+Reported causal extraction improvement
5Core research areas
Questions worth
solving.
My work explores how AI can become more reliable, explainable, resilient, and useful in high-stakes environments.
02
Causal AI & NLP
Extracting causal relationships from complex technical, scientific, medical, and incident-report text using fine-tuned language models.
T5BERTNLP
03
Trustworthy AI
Studying fairness, explainability, bias mitigation, and evaluation strategies for intelligent systems used in sensitive domains.
XAIFairnessEthics
04
Robust Healthcare AI
Multimodal clinical decision support and adversarially robust models combining medical imaging with clinical language.
ViTBioBERTPGD
05
Secure Intelligent Systems
Resilient audio classification, backdoor attacks, IoT security, and adaptive communications for next-generation networks.
IoT6GSecurity
Research,
in print.
A growing collection of peer-reviewed work and thesis research.
Research in
practice.
Graduate Research Assistant
University of Nevada–Las Vegas · Department of Computer Science
Developed multimodal glaucoma detection and segmentation models using fundus imaging, clinical notes, Vision Transformers, and BioBERT. Reported a 12% improvement in diagnostic accuracy and AUC > 0.9 on clinical datasets.
Graduate Research Assistant
University of Idaho · MIDA Lab
Pioneered a causality extraction framework using fine-tuned T5 and BERT models; reported more than 20% improvement in root-cause detection in nuclear licensee event reports. Designed a hybrid token/sequence-level loss.
Graduate Teaching Assistant
University of Idaho · Department of Computer Science
Supported Python for Machine Learning and core computer science instruction, mentoring more than 120 students across data structures, algorithms, OOP, and programming labs.
Academic
foundation.
M.Sc. Information Studies
Trine University
GPA 4.0
M.Sc. Computer Science
University of Idaho
GPA 4.0
B.Eng. Software Engineering
Sichuan University
GPA 3.93Tools I
work with.
PythonPyTorchTensorFlowHuggingFaceTransformersOpenCVScikit-learnCUDADockerAWSGCPAzureSQLApache SparkPandasNumPyJavaC/C++RGitLinuxLaTeXJupyterREST APIs
Recognition
earned.
Top 5 PerformerDeep Learning for Medical Imaging · University of Idaho, Moscow, USA · Recognized for developing high-performing AI models for medical image classification · 2024
6th Place · Programming ContestDaffodil International University, Dhaka, Bangladesh · Ranked 6th among 319 participants · 2017
One Belt One Road ScholarFull scholarship recipient · Sichuan University · 2018–2021
Outstanding Class CadreRecognized for student leadership and community service · 2021
Let's build something
meaningful.
Research collaboration, academic opportunities, or questions about my work — I'd be glad to connect.
Email ↗