Lecturer in Data Science and Artificial Intelligence at SRH University, and PhD graduate from the Faculty of Mathematics and Computer Science, Heidelberg University.

My research focuses on natural language processing for biomedical text, with an emphasis on class imbalance and cost-sensitive fine-tuning of pre-trained language models. I've learned as much from failed experiments as from successful ones, and I try to carry that into how I teach.

I value academia deeply. Lecturing, sharing knowledge, and collaborating with others are what energise me. I am eager to pursue opportunities where research and teaching intersect, and where I can contribute while continuing to grow.

• Natural Language Processing (NLP) for biomedical text
• Joint Named entity recognition & relation extraction
• Class-imbalance problem in NLP
• Cost-sensitive fine-tuning of pre-trained language models

• Lecturer, Data Science and Artificial Intelligence: SRH University of Applied Sciences, 2025–present
• PhD, Mathematics and Computer Science (Data Science): Heidelberg University, 2020–2025
• Research Associate, Scientific Databases and Visualization: Heidelberg Institute for Theoretical Studies (HITS gGmbH), 2020–2025

As a Lecturer at the Department of Applied Data Science and Artificial Intelligence, I teach courses including Python Foundations for Data Science, Advanced Python for Data Science, Statistics and Machine Learning, Natural Language Processing, and Generative AI. I supervise Master's research projects from conception to submission and collaborate on AI research projects with industry and academic partners.

• WeLT → A cost-sensitive BERT that handles class imbalance for biomedical NER
• WeLT-impact-on-BioNEL → Studying WeLT's impact on biomedical named entity linking
• WeLT-ASpERT / WeLT-SpERT / WeLT-TablERT-CNN → Joint entity and relation extraction trainers built on the WeLT framework

I've published 280+ fine-tuned models on the Hugging Face Hub, covering token classification (NER) across biomedical benchmarks such as NCBI-disease and BC5CDR, built on backbones like SciBERT, BlueBERT, and BioELECTRA with application of the WeLT cost-sensitive training approach.

I've had the honour of reviewing manuscripts for Database (Oxford University Press) since 2024. Engaging in the peer review process is a valuable way to give back to the scientific community and uphold research integrity.

• Academic collaborations and joint research
• Guest lectures and workshops
• PhD/Master's co-supervision
• Peer review and editorial contributions

• Mobasher, G. (2025). WeLT: Weighted Loss Trainer for Biomedical Joint Entity and Relation Extraction (Doctoral dissertation). Heidelberg University.
• Mobasher, G., Müller, W., Krebs, O., & Gertz, M. (2023). WeLT: Improving Biomedical Fine-tuned Pre-trained Language Models with Cost-sensitive Learning. 22nd Workshop on Biomedical Natural Language Processing (BioNLP).
• Leaman, R., Islamaj, R., et al., incl. Mobasher, G. (2023). Chemical identification and indexing in full-text articles: an overview of the NLM-Chem track at BioCreative VII. Database, 2023, baad005.
• Mobasher, G., Mertová, L., Ghosh, S., Krebs, O., Heinlein, B., & Müller, W. (2021). Combining dictionary- and rule-based approximate entity linking with tuned BioBERT. bioRxiv.