Johanna Björklund

This PhD project at Umeå University sits in the Department of Computing Science and focuses on machine learning for graph transformations . The research aims to combine deep learning with grammatical inference to build models that are interpretable, efficient, and mathematically verifiable, while still benefiting from the flexibility of neural methods. The project is centered on graphs as a representation of structured information, from sentences and computer programs to social networks, and on how graph-based methods can help make future generative AI systems safer and more reliable. Rather than relying on opaque heavy models, the work explores transparent, rule-based rewriting systems that can transform graphs, validate AI-generated structures, and support inference-time guidance for neural networks. The student will investigate learning approaches that draw on grammatical inference , reinforcement learning , and efficient search algorithms to discover both rules and probabilities for complex graph behavior. The project is intended to produce research relevant to theoretical computer science and symbolic machine learning , with possible implications for automated theorem proving, knowledge-graph inference, and causal analysis. The PhD candidate will work closely with the PI and her research group in a research environment described as inclusive, collaborative, and bottom-up driven. The department highlights rapid growth, a diverse international community, and strong support for career development, networking, and administrative and technical services.

Umeå University, one of Sweden’s largest and most dynamic higher education institutions, is renowned for its world-leading research and innovative academic environment. The university’s Department of Computing Science is seeking a PhD student for a joint position with the Umeå Centre for Gender Studies, focusing on the development of AI methods for multilingual gender bias editing. This project aims to address the detection and mitigation of gender biases in language technologies, particularly within large language models and generative AI systems, which can perpetuate social harms and inequalities. The research will develop culturally informed tools for bias detection in NLP models, capable of operating across diverse cultural contexts and psychosocial factors. The project is especially challenging due to the context-dependent nature of language, ambiguity, and the variability of bias signals across communities. It places particular emphasis on low- and medium-resource languages, where data scarcity complicates both measurement and mitigation. By collaborating with experts in gender studies and linguistics, the successful candidate will help design adaptable tools for gender-bias detection and editing, contributing to the broader goal of promoting social good and equity in AI applications. As a PhD student, you will join a vibrant, inclusive, and interdisciplinary research environment, benefiting from career development support, networking opportunities, and excellent employment conditions. The position is primarily devoted to research (at least 80%), with the possibility of up to 20% departmental service, such as teaching, which can extend the position to a maximum of five years. The employment includes a competitive salary according to the university’s established doctoral student salary ladder. Eligibility requirements include a second-cycle degree or at least 240 ECTS credits (with at least 60 ECTS at second-cycle level), or equivalent international qualifications. Applicants must have completed at least 90 higher education credits in computer science or closely related subjects. Experience in Natural Language Processing is essential, along with a strong interest in Gender Studies and interdisciplinary research involving psychology and the social sciences. Additional merits include competence in Gender Studies, proficiency in low- to medium-resourced languages, experience in multilingual NLP, multimodal AI applications, and interdisciplinary research. The department values diversity and particularly encourages women and those outside the gender binary to apply. Applications must be submitted electronically via the Umeå University e-recruitment system by 2026-01-30. Required documents include a cover letter, CV, copies of theses and publications, degree certificates, documentation of relevant experiences, contact information for two references, a one-page research plan, and a half-page statement on Gender Studies experience. All documents should be in PDF format and written in English or Swedish. For further information, prospective applicants may contact Associate Professor Henrik Björklund, Professor Johanna Björklund, or Associate Professor Loïs Vanhée. This position offers a unique opportunity to contribute to cutting-edge research at the intersection of AI, language technology, and gender studies, within a supportive and forward-thinking academic community.