1-3 Postdoctoral Researchers, Project Researchers or Research Assistants for Training Large Language Models

Join Europe's leading experts in creating the next generation of European generative AI models!

TurkuNLP (https://turkunlp.org) is an internationally recognized multidisciplinary research group specializing in the development of machine learning-based resources, models, and tools for natural language processing (NLP) tasks using internet-scale data. In recent years, TurkuNLP has created several openly available Transformer-based language models, such as FinBERT and FinGPT, Finland's leading BERT and GPT-3 models, and subsequently the Poro and Viking models. TurkuNLP was also selected as a pilot user of Europe's largest supercomputer, LUMI, and the group has a long tradition of using high-performance computing in NLP.

TurkuNLP is seeking 1-3 postdoctoral researchers, project researchers, or secondarily research assistants for positions involving the training of large language models. The work begins as soon as possible by mutual agreement. The positions are fixed-term and will last for nearly 3 years until January 31, 2028 or February 28, 2028 depending on the project. The work can be done either full-time or part-time, depending on the applicants' wishes and possibilities. The positions are part of the EU-funded High Performance Language Technologies (HPLT) research project, as well as the follow-up projects LLMs4EU (end date 28.2.2028) and OpenEuroLLM (end date 31.1.2028). These projects aim to create a new generation of large language models that cover all European languages. These models will be fine-tuned using terabytes of text data and Europe's largest supercomputer, LUMI.

More information https://hplt-project.org/, https://openeurollm.eu/

Job description

This position offers an excellent opportunity to participate in the development of large language models within a research group that possesses top-tier expertise, data, and computational resources. The projects differ from many similar projects in that (1) they have access to millions of GPU hours of computing time, (2) they have access to extremely large text datasets, including the Internet Archive dataset, and (3) they are multinational collaborations that combine the NLP expertise of several research groups and industry companies. The projects are well networked and maintain regular interaction with key players in the field. As part of the project, it is possible, but not mandatory, to also pursue a doctoral degree under a separate agreement. If you are interested in this opportunity, please mention it in your motivation letter.

Key tasks and responsibilities

You will work as part of the TurkuNLP language model project team. The key tasks involve the training and evaluation of very large multilingual language models from the ground up. The project focuses particularly on training models on the LUMI supercomputer's AMD architecture. Postdoctoral researchers are also expected to participate in the supervision of junior researchers and other similar tasks typically associated with the responsibilities of doctoral graduates in academia.