Muhammad Helmy

NLP is where I combine my passion for mathematics, coding, and linguistics, particularly focusing on its applications in law and religious studies. I developed strong programming skills while competing in the ICPC, and built solid foundations in mathematics and machine learning through studying rigorous books and courses. I have contributed to several NLP projects and ranked 1st in the AraGenEval shared task of the ArabicNLP 2025 conference, and was recently selected for the competitive Fatima Fellowship to conduct research on reinforcement learning.

Research Fellow at Fatima Fellowship (2026-05 – Present)

• Research Project: Exploration Dynamics in Policy Gradient Methods for Multi-Armed Bandits
• Mentor: Prof. Yassir Jedra (Imperial College London)

NLP Researcher at ArabicNLP 2025 Conference (2025-07 – 2025-09)

Participated in the AraGenEval shared task at the ArabicNLP 2025 Conference, ranked 1st.

• Developed an ensemble system by fine-tuning and integrating 4 transformer-based models to identify the author of a given text excerpt
• Our team achieved a Macro-F1 score of 90%, representing a 15% improvement over the baseline

Competitor at Maqsam PII Redaction Competition (2025-06 – 2025-08)

Participated in the Maqsam PII Redaction Competition, ranked 3rd place.

• Designed preprocessing pipelines and prompt-based templates for source-target formatting to anonymize sensitive information in Arabic text
• Fine-tuned Small Language Models using QLoRA via the Unsloth framework on Kaggle T4 x2 GPUs

Bachelor in Computer Science – Nile University (2020-09 – 2024-06)