About

I am an experienced machine learning engineer with a background in NLP research. I’m currently building on-prem AI tools at Migros Bank in Zurich, Switzerland.

Before that, I worked on serving LLMs and NMT models at the language service provider Supertext, which emerged from a merger with the machine translation startup Textshuttle that I had joined in early 2024.

I did my graduate studies in computer science at McGill University and the Mila Quebec AI Institute in Montreal. During that time, I did research under the supervision of Jackie C. K. Cheung in collaboration with Microsoft Research on natural language understanding issues such as knowledge representation and common-sense reasoning.

In my undergrad, I studied cognitive science at Osnabrück University in Germany and Chuo University in Japan. In my leisure I enjoy playing Go, reading novels, and spending time in nature. You can find some of my hiking and snowshoe tours on this website as well.  

Publications

Alex Flückiger, Chantal Amrhein, Tim Graf, Frédéric Odermatt, Martin Pömsl, Philippe Schläpfer, Florian Schottmann, Samuel Läubli (2025). A comparison of translation performance between DeepL and Supertext. Proceedings of Machine Translation Summit XX: Volume 2. Geneva, Switzerland: European Association for Machine Translation. Data & Code.

Akshatha Arodi*, Martin Pömsl*, Kaheer Suleman, Adam Trischler, Alexandra Olteanu, and Jackie Chi Kit Cheung (2023). The KITMUS Test: Evaluating Knowledge Integration from Multiple Sources. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Toronto, Canada: Association for Computational Linguistics. Nominated for a Best Paper Award. *Equal contribution. Data & Code.

Martin Pömsl and Roman Lyapin (2020). CIRCE at SemEval-2020 Task 1: Ensembling Context-Free and Context-Dependent Word Representations. In Proceedings of the Fourteenth Workshop on Semantic Evaluation. Barcelona (online): International Committee for Computational Linguistics, pp. 180–186. Data & Code.

Theses

Martin Pömsl (2023). Evaluating the Integration of Pretrain-Time and Inference-Time Knowledge in LLM-based NLU Systems. Master Thesis. McGill University.

Martin Pömsl (2021). Extending Neural Temporal Tagging Systems with External Knowledge. Bachelor Thesis. Osnabrück University.