TurkuNLP (https://turkunlp.org) is an internationally established multidisciplinary research group that specializes in the development of machine learning -based resources, models, and tools for natural language processing (NLP) using web-scale datasets. In recent years TurkuNLP has created a number of openly available Transformer-based language models, including e.g. FinBERT and FinGPT, the leading Finnish BERT and GPT-3 models. TurkuNLP was also selected as the pilot user of the Europe’s largest supercomputer LUMI and has a long tradition in the use of high performance computing in NLP.
We welcome applications for 1 fixed-term position of a Postdoctoral Researcher to work on large language model pre-training. The position is for the period of February 1, 2024 (a more precise date can be agreed upon) to August 31, 2025, and can be either part-time or full-time accommodating the applicants’ wishes and availability.
The position is part of the research project High Performance Language Models (HPLT) funded by Horizon Europe. The project will create next-generation large language models covering all European languages and more by training on terabytes of data using the largest supercomputer in Europe, LUMI. For more information, please visit https://hplt-project.org/ .
We offer you
The position offers an excellent opportunity to participate in the development of large language models in a team which has the experience, data, and computational resources. The HPLT project is set apart from many similar efforts by (1) having a very large allotment of computational time at its disposal, on the order of 15+ million GPU hours on Europe’s largest supercomputer, (2) having secured access to very large textual datasets, including partial dumps of the Internet Archive, and (3) being a multinational collaboration, bringing together NLP expertise from several well established labs and data companies. The project is well networked and regularly interacts with the key players in the field.
Key tasks and responsibilities
You will work as part of the HPLT project team in TurkuNLP. The key tasks will revolve around the steps needed to pre-train, from scratch, very large multilingual language models. A particular focus will be on training large multi-lingual models on the AMD architecture of the LUMI supercomputer. Post-doctoral researchers are additionally expected to participate in the supervision of more junior researchers and other similar tasks typically associated with a post-doctoral position in academia.
Turku region with its 320,000 people is a major urban area and a leading hub for technological development and economic growth in Finland. This former capital, located in the southwestern Finland, ...
The University of Turku has a unique, creative and inspirational work environment. Here you will work with top experts, pedagogues and researchers.
Visit the employer page