Clinical NLP · LLM evaluation · Medical dialogue

I'm an MSc student in Natural Language Processing at MBZUAI and a member of HALO Lab, co-supervised by Dr. Xiuying Chen and Dr. Imran Razzak. I build benchmarks and methods for the parts of clinical work that exam-style datasets skip: multi-turn consultations, patients describing symptoms in their own words, noisy transcripts and handwritten prescriptions.

I'm currently a research intern at DFKI Kaiserslautern with Dr. Sheraz Ahmed, and before that I worked at Microsoft, Botit and Monta.ai

News

2026.09Started a research internship at DFKI Kaiserslautern, working on clinical information extraction and medication decision support.
2026.09MedRoundsQA is on arXiv, with code and data on GitHub.
2026.08MedRoundsQA accepted to the EMNLP 2026 Main Conference. See you at the poster in Budapest.
2026.05The Cylindrical Representation Hypothesis for LLM steering, with HALO Lab, accepted to ICML 2026.
2025Joined MBZUAI for an MSc in NLP. DA-MTL published at SecureComm 2025.

Research interests

01

Interactive medical dialogue

How LLM doctors gather evidence over many turns, and how fairly they treat patients who communicate differently.

02

Clinical information extraction

Treatment plans, medications and medication changes from consultations, notes and patient-written text.

03

Vision-language models in medicine

Reading real handwritten prescriptions, and how VLMs reason over biomedical imagery.

04

Arabic & multilingual NLP

Arabic text-to-SQL, document understanding, and accent-diverse speech for Arabic medical dialogue.

Selected publications

Full list on Google Scholar
EMNLP 2026 · MainFirst author

MedRoundsQA: A Persona and Difficulty Aware Evaluation for Multi-Turn Medical Consultations

Youssef Mohamed, Ahmed Heakl, Qinrong Cui, Junhong Liang, Rafiq Ali, Bdour Babillie, Nazira Dunbayeva, Lang Gao, Omar Hussein, Ahmed Nada, Ahmed Mohamed Magdy Mohamed, Jinghui Liu, Salman Khan, Imran Razzak, Yuxia Wang, Xiuying Chen

Same case, same facts, different patient: moving from single-turn diagnosis to a real consultation costs 15 LLMs 13–39 points, and how educated the patient sounds shifts accuracy by another 7–8.

PaperCode & data

ICML 2026

The Cylindrical Representation Hypothesis for Language Model Steering

Lang Gao, Jinghui Zhang, Wei Liu, Fengxian Ji, Chenxi Wang, Zirui Song, Akash Ghosh, Youssef Mohamed, Preslav Nakov, Xiuying Chen

Relaxing the orthogonality assumption of the Linear Representation Hypothesis explains why steering directions interact and destabilise model behaviour.

PaperCode

SecureComm 2025

Two Birds with One Stone: Multi-Task Detection and Attribution of LLM-Generated Text

Zixin Rao*, Youssef Mohamed*, Shang Liu, Zeyan Liu (*equal contribution)

A multitask model that both detects machine-generated text and attributes it to its source LLM, across languages.

Paper

ACLing 2024

ResuméAtlas: Revisiting Resume Classification with Large-Scale Datasets and Large Language Models

Ahmed Heakl*, Youssef Mohamed*, Noran Mohamed*, Aly Elsharkawy, Ahmed Zaky (*equal contribution)

PaperCodeDataset

Experience & education

2026 – now
Research Intern · DFKI Kaiserslautern
With Dr. Sheraz Ahmed. Clinical information extraction, medication decision support, VLMs for biomedical imaging.
2025 – 2027
MSc, Natural Language Processing · MBZUAI, HALO Lab
Thesis on NLP for healthcare. Co-supervised by Dr. Xiuying Chen and Dr. Imran Razzak.
2025
Machine Learning Researcher · Monta.ai
Benchmarks for Arabic document layout detection and figure understanding.
2024 – 2025
Visiting Research Student · UNSW Sydney
Multitask learning for medical image classification and retrieval.
2024 – 2025
AI Research Intern · University of Louisville
Arabic language analysis and multilingual LLM-generated text detection.
2024 – 2025
Machine Learning Researcher · Botit
Transformer-based semantic search for e-commerce.
2023
Applied Data Scientist Intern · Microsoft
Transformer-based emoji prediction (+7% top-3 F1) and deployment tooling.