Elias Hossain

I am a Ph.D. student at the University of Central Florida , advised by Dr. Niloofar Yousefi. My research advances Large Language Models (LLMs) and Machine Learning for biomedical discovery and antimicrobial resistance (AMR) prediction. I am developing an AI-driven framework that integrates LLM-based reasoning with computational biology to predict resistance mechanisms and design digital twin systems capable of simulating host–pathogen interactions for drug discovery in bacterial pathogens. My work aims to enable trustworthy, interpretable, and data-efficient biomedical AI that bridges computational intelligence with clinical translation. With a background as a Software Engineer and ML Researcher, I bring expertise in backend system design, algorithm optimization, and scalable AI model deployment, driving robust and explainable AI solutions for healthcare and life sciences.

Education

Research Spotlight

My current research spans reinforcement learning for pathogen evolution, Bayesian uncertainty in NLP, preference alignment robustness, and tokenization theory for domain-specific language models. Below are selected ongoing and recently submitted works.

Paper 01

⏳ Under Review — ICML 2026

Modeling Antimicrobial Resistance Evolution as a Biologically Constrained Sequential Decision Process

AMR evolution is modeled as a reinforcement learning problem over a digital twin environment, using a GCN encoder over multi-relational biological interaction graphs. The agent learns adaptive, trajectory-level resistance strategies — mutation, HGT, and neutral maintenance — revealing evolutionary dynamics beyond static resistance prediction.

Reinforcement LearningDigital Twin AMRGraph Neural Networks Computational Biology

arXiv — Coming Soon

Paper 02

⏳ Under Review — ACL ARR 2026

UAT-Lite: Uncertainty-Aware Attention for Pretrained Transformers via Epistemic Modulation

An inference-time method that uses epistemic uncertainty from Monte Carlo dropout to modulate self-attention weights in pretrained transformers. Improves calibration, selective prediction, and robustness across NLI, QA, and sentiment tasks — without any retraining.

Bayesian NLPUncertainty Quantification MC DropoutTransformers Calibration

arXiv

Paper 03

ICLR Trustworthy AI Workshop 2026

SAVe-DPO: Multi-Agent Verification for Robust Preference-Based Alignment

Proposes a multi-agent verification layer — comprising a Knowledge Verifier, Behavior Auditor, Ethics Evaluator, and Trust Assessor — to filter corrupted preference supervision prior to DPO/RLHF training, yielding more robust and trustworthy aligned language models.

RLHF / DPOAlignment Multi-AgentTrustworthy AI LLM Safety

arXiv — Coming Soon

More on Google Scholar  ·  Medium

Selected Publications

• 1
  Hossain, E. "TopNet R1: A Multi-Stage AI Framework for Topic Discovery in Scientific Abstracts." Master's Thesis, Mississippi State University, 2025. Thesis
• 2
  Hossain, E., Saha, S., Roy, S., and Prasad, R. "Can Transformer Memory Be Corrupted? Investigating Cache-Side Vulnerabilities in Large Language Models." arXiv preprint, 2025. arXiv
• 3
  Hossain, E., Nuzhat, T., Masum, S., Rahimi, S., and Golilarz, N. A. "R-GAT: cancer document classification leveraging graph-based residual network for scenarios with limited data." Nature Scientific Reports, 2026. Nature
• 4
  Hossain, E., Hasan Nipu, M. M., Sheikh, M., Rana, R., Neupane, S., and Yousefi, N. "MedBayes-Lite: Bayesian Uncertainty Quantification for Safe Clinical Decision Support." arXiv
• 5
  Hossain, E., Rana, R., et al. "Natural Language Processing in Electronic Health Records for Healthcare Decision-Making: A Systematic Review." Computers in Biology and Medicine, vol. 155, 2023. DOI

Beyond Research

Science consumes most of my hours but the hours outside of it are just as defining. I write fiction that explores human connection and quiet resilience, and I spend my evenings on the tennis court, chasing the same focus and composure I try to bring to research. These aren't distractions from my work; they're the reason it has depth.

Literary Fiction  ·  Romance

Through Her Lens — Vol. I

The Shore Between Us

by Elias Hossain  ·  First Edition, 2025

Arthur Hayes is a photographer who retreated to the quiet shores of Crescent Bay to outrun a broken engagement and two years of silence. His craft is technically flawless, yet his frames are conspicuously empty of people. Then his neighbor's cottage, long vacant, gets a new tenant: Isla, who carries her own carefully guarded grief. Two people who built walls out of routine and solitude meet in one small, salt-weathered bay where those walls begin to crack. The Shore Between Us is a slow-burn story about what happens when two people who have forgotten how to be seen finally, reluctantly, let each other look.

"For everyone who's ever had to rebuild themselves from scratch. May you find your own shore, and someone to cross it with."

❤️ This Book Is Written for a Purpose, Not a Career

I didn't write this to build an author brand. I wrote it because every cent this book earns goes entirely to charity. There is no personal income and no profit kept. This story exists so that people who need help the most can benefit from every copy you read.

Charitable Giving Humanitarian Support Education & Food Access

100% transparent. I will publicly share total sales, earnings, and every charity disbursement right here — so you can always see exactly where your money goes and the real impact it creates.

Get the Book & Give Back Releasing on Amazon soon

Tennis taught me the same things a PhD does — that consistency beats brilliance, that patience under pressure is a learnable skill, and that losing a set is never the end of the match. Whether it's an early-morning rally before seminars or a late-night session under the floodlights, the court is where I reset.

Evening session

Golden hour on court

Night match

Late-night rally under the lights

Contact