Hello, my name is
Md. Shahidul Salim
PhD Student at University of Massachusetts Lowell, USA
🤗

My research focuses on developing reliable, clinically grounded language technologies for healthcare. I work on medical LLM safety and robustness, controlled clinical text generation, GLP-1 discontinuation reason classification, and multilingual medical translation. My current research explores reinforcement learning, multi-agent debate, knowledge augmentation, and uncertainty quantification to improve the reliability of AI systems in real-world clinical settings.

🍀 If you are interested in collaboration or research opportunities, feel free to reach out!

Get in Touch
Md. Shahidul Salim

Recent Updates

About Me

I am a PhD student and Research Assistant at the University of Massachusetts Lowell, where I work under the supervision of Dr. Hong Yu. I am currently on study leave from my position as a Lecturer in the Department of Computer Science and Engineering at Khulna University of Engineering & Technology (KUET).

My research centers on building safe, robust, and clinically grounded language technologies for healthcare. I study medical LLM behavior under challenging patient interactions, readability-controlled clinical summarization, GLP-1 discontinuation reason classification, and multilingual medical translation. My work draws on reinforcement learning, multi-agent debate, knowledge augmentation, and uncertainty quantification to improve the reliability and usefulness of clinical AI systems.

During Summer 2026, I collaborated with Dr. Hong Yu and the VA research team at the Edith Nourse Rogers Memorial Veterans Hospital in Bedford, Massachusetts. I applied machine learning and deep learning methods to support NLP research, remotely analyzed VA research data, and maintained patient confidentiality and HIPAA compliance. I have also served as a reviewer for EMNLP 2026 and an external reviewer for AAAI 2024.

Academic & Work Timeline

Research Interests

Medical LLM Safety & Robustness Challenging Patient Behaviors Controlled Clinical Generation GLP-1 Discontinuation Classification Multilingual Medical Translation Multi-Agent Debate Reinforcement Learning Knowledge Augmentation Uncertainty Quantification Medical NLP

Hobbies

Outside of research, I enjoy playing FIFA, traveling, and watching sci-fi and thriller movies. Some favorites include Interstellar, Dark, and Stranger Things.

Research Areas

Medical LLM Safety & Robustness

Evaluating and improving medical LLMs under challenging patient behaviors using reinforcement learning, multi-agent debate, guardrail methods, and uncertainty analysis.

Clinical NLP & Controlled Generation

Building clinically grounded methods for readability-controlled summarization and classifying GLP-1 discontinuation reasons from clinical notes.

Multilingual Medical NLP

Studying when dictionary and knowledge augmentation improve multilingual medical translation, with an emphasis on accuracy, efficiency, and clinical reliability.

🏆 Selected Research Papers

Natural Language Processing

    Machine Learning / Deep Learning

      📚 Other Publications

      Natural Language Processing

        Machine Learning / Deep Learning

          🔍 Under Review

            🔬 Ongoing Research

            Submitted to EACL on August 3

            1. A Multilingual Evaluation of Medical LLM Robustness Under Challenging Patient Behaviors - I completed the project and submitted it to EACL.
            2. ReadAlign: Learning Clinically Grounded Style Signals for Readability-Controlled Clinical Summarization - Following the reviewers' feedback, I completed the remaining experiments, incorporated the additional results into the revised work, and submitted it to EACL on August 3.
            3. GLP-1 Discontinuation Reason Classification: A Controlled Clinical NLP Benchmark from Synthetic Notes - Adiba and I completed this project and submitted it to EACL on August 3.

            Currently Working On

            1. When Does Dictionary Augmentation Help Multilingual Medical Translation?
            2. Improving model robustness to challenging patient behaviors using multi-agent debate.

            Broader Research Directions

              💻 Projects

              My Skills

              Medical AI & LLM Engineering

              I specialize in designing, adapting, and evaluating large language models for clinical NLP. My experience includes parameter-efficient fine-tuning, reinforcement learning, prompt and knowledge augmentation, multi-agent workflows, synthetic clinical data generation, and uncertainty-aware model evaluation. I build research pipelines with PyTorch, Hugging Face Transformers, VerL, Unsloth AI, and vLLM.

              I am proficient in Python and C/C++ and experienced with model serving, retrieval, and reproducible experimentation using FastAPI, Docker, FAISS/Chroma, Weights & Biases, Git, and Linux. I also use LangChain, LlamaIndex, and Gradio to prototype and deploy LLM-based applications.

              Detailed Experience
              PyTorch VerL Unsloth AI vLLM Hugging Face Transformers LangChain LlamaIndex PEFT Reinforcement Learning LLM Evaluation Multi-Agent Systems Prompt Engineering Knowledge Augmentation Uncertainty Quantification Synthetic Data Generation Medical NLP Clinical Summarization Machine Translation Gradio FAISS/Chroma Weights & Biases Python C/C++ FastAPI Docker Linux Git NumPy pandas LaTeX HTML/CSS

              Contact Me

              Email

              mdshahidul_salim@student.uml.edu

              shahidulshakib034@gmail.com

              Address

              132 University Ave, Lowell, MA, USA

              Social Media