ML engineer & researcher, M. Sc.
I build NLP models and the data infrastructure behind them, from embedding models trained on multi-GPU nodes to pipelines over hundreds of millions of papers.
Finishing my PhD at TU Hamburg in 2027 and looking for industry roles in applied ML, NLP and R&D.
2026
Embedding model for scientific documents, trained with the sentences in which papers cite each other as contrastive signal. ModernBERT-base with up to 8K context; six checkpoints released under MIT.
Up to ~30M training pairs on 4×H200, SciRepEval 66.85 (EMNLP 2026)
PyTorch, ModernBERT, Contrastive learning, DuckDB
2026
Locally deployable infrastructure that joins eight scholarly databases (Semantic Scholar, OpenAlex, Crossref and others) into one SQL interface, with embedding-based alignment of 4,516 topics to 13 ontologies.
293M papers, ~960 GB of Parquet behind 153 DuckDB views
Python, DuckDB, Parquet, Embeddings
2026
Jonas Wilinski is an ML engineer and PhD candidate at the Hamburg University of Technology (TUHH). He trains NLP models and builds the data systems they need, and uses them to study how AI concepts spread through science. Before the PhD he worked as CTO of the supply chain start-up Cargofaces, in digitalization at Porsche AG, and as a freelance IT consultant; the details are on his experience page. He holds an M.Sc. in International Industrial Engineering from TU Hamburg and a B.Sc. in Electrical Engineering & IT from Kiel University. His publications include work at EMNLP and DRUID. His stack spans Python, PyTorch, SQL and DuckDB, Docker, React and Swift. Outside work he sails and films his trips; some of them are on his hobbies page.
Roles in applied ML and R&D, collaborations, or a question about one of my projects: email is the fastest way to reach me.