mohammad gorzin

Engineering and deployment of a German NLP pipeline for automatic coreference resolution in literary documents. The project focused on transforming an existing research workflow into a containerized GPU-based AI service and extending it with LLM-generated character knowledge and embeddings. • Containerized and deployed the transformer-based coreference resolution pipeline using Docker, Kubernetes, and GPU inference. • Developed a FastAPI service covering preprocessing, model inference, postprocessing, and g

Provided software and hardware support, troubleshooting system and user issues in an academic environment. Assisted with device setup, installations, and maintenance for staff and students

I had the opportunity to contribute significantly to citation lifecycle analysis at CAIDAS, University of Würzburg. • Built a citation analysis pipeline that examines how citation intent evolves over time. • Developed an LLM-Council framework for creating silver-standard citation intent annotations. • Analyzed citation behavior trends through lifecycle visualizations, enhancing understanding of landmark publications.

• Built a large-scale citation analysis pipeline to study how citation intent evolves throughout the lifecycle of landmark scientific publications. • Collected and processed more than 17,000 citation contexts from Semantic Scholar and unarXive datasets. • Implemented citation intent classification pipelines using DSPy, large language models, and the ACL-ARC benchmark. • Developed an LLM-Council framework for creating silver-standard citation intent annotations. • Analyzed temporal citation behavior of the W

• Built multivariate time-series representations from literary narratives using sentiment trajectories, danger scores, and scene-level features. • Trained and evaluated the MOMENT foundation model for happy-ending prediction and genre classification tasks. • Conducted hyperparameter optimization, feature engineering, and validation studies on datasets containing more than 200 novels. • Developed reproducible GPU-based experiment pipelines on HPC infrastruc- ture using SLURM. • Investigated the impact of nar

At WueDive - Digital Teaching Hub, I focused on automating virtual microscopy image processing. I engineered a Python solution that converted large datasets from Mrxs to PNG format, ensuring superior image quality through lossless compression. By integrating with Zoomify, I facilitated interactive microscopy experiences, completing all slide conversions in just two days.

• Developed advanced search and recommendation features utilizing Elasticsearch and user-history signals. • Enhanced search accuracy and user engagement through data-driven feature development. • Collaborated with cross-functional teams to align product features with user needs.

• Managed and optimized MySQL databases to ensure data integrity and efficient performance. • Developed and implemented API access management logic using PHP and Laravel, improving security and streamlining data access for third-party applications. • Collaborated with the UI team to develop and enhance web pages, contributing to a 10% increase in user satisfaction based on website feedback surveys.