Tillbaka till alla jobb

Master's Thesis Workers, Applied LLMs and Multimodality

VTT · Espoo

Fler jobb hos VTT (13) →
Hybridtrainee💰 ~3 400–4 150 EUR uppsk.IT och teknik · ForskningPublicerad 11.11.2026 02.00

Kompetenser

aiMachine Learningnatural language processingmultimodal airesearch

Jobbeskrivning

Are you all about science and technology? Would you like to work in a meaningful role with real responsibility and excellent learning opportunities? Do you want to work with superb colleagues who are experts in their field to drive a sustainable future and tackle the greatest challenges of our time? VTT might be just the place for you. VTT is one of the leading applied research organizations in Europe. What unites us are curiosity, a passion for learning and a purpose: to solve the biggest challenges of our time and turn them into sustainable growth and well-being. Our cross-disciplinary community of over 2000 brilliant minds is focused on creating impact through science-based innovation. Get to know us further by exploring VTT's tech & infra . We are now looking for two Master’s Thesis Workers to join our Reliable Intelligence team in the Physical AI unit. Join us - and let’s make an impact together! You will do your thesis in Excellent-AI, an international ITEA research project that helps European businesses become AI-first organizations. Together with companies and research partners from several countries, the project develops AI-based tools for strategic leadership, innovation and governance, so that organizations can adopt AI in a way that is effective, ethical and compliant. As a thesis worker, you will do hands-on AI development for a use case, so you will work with real data, requirements, and company partners. Each of the two positions has its own focus: Large language models and natural language processing. You will work in depth with LLMs, going beyond using them as ready-made tools. Your research will target core NLP problems behind the use case, such as improving text and information extraction, calibrating model confidence so that a system knows when its output should be checked by a person, and resolving matches between records that describe the same thing differently. Rule-based matching only succeeds when everything agrees exactly, and most of the remaining manual work comes from ordinary human variation: a typo in a product code, the same item described in different words or languages, or quantities given in different units. You will study how LLMs can resolve these cases reliably. Multimodal AI and agents. You will work in depth with two fast-moving areas and the connection between them. On the multimodal side, your research will target problems such as multimodal retrieval, where a system searches text, images and figures together, and keeping model outputs grounded in what the input actually shows. On the agent side, you will study how several AI agents can be orchestrated to plan and carry out complex, multi-step workflows together with the professionals they support: how to divide work between agents, how to stop small errors from compounding over many steps, and when an agent should hand a decision back to a person. Bringing the two together, you will explore how agents can reliably use information that comes in many forms. The exact scope of your thesis is not fixed. We want to hear which challenges and technologies interest you most, and we will shape the topic together with you. What unites both positions is our research focus: methods that reduce hallucinations and increase the reliability of AI, so that its outputs can be trusted in business use. In the Reliable Intelligence team, part of VTT’s Physical AI unit, we work to make AI-enabled technology and decisions reliable in real-world use for industry and society. You will be supervised by experienced researchers and learn how AI research is done in practice, from a first idea and experiment to results that a company can use and the research community can build on. As Master’s Thesis Worker, you will: Develop AI solutions for a use case defined with a company partner. Review the state of the art in your topic and identify where current methods fall short. Design, implement and test novel methods, and compare them against strong baselines. Prepare and analyse data, and design evaluation setups that measure quality, robustness and reliability. Document your experiments and results, and present them to the project team and partners. Write your Master’s thesis on the work, with the aim of contributing to a scientific publication. Who we are looking for: You are a Master’s student in computer science, data science, machine learning, or a related field, ready to start your thesis. You have solid programming skills in Python and experience with machine learning libraries (e.g. PyTorch, Hugging Face). You have a good understanding of machine learning and deep learning, including how modern neural language or vision models work. You have a particular interest in either NLP and language models, or multimodal models and AI agents. You are motivated by research: you enjoy reading scientific papers, trying out new ideas and evaluating them rigorously. Hobby or study projects related to fine-tuning or evaluating LLMs, information extract
Logga in gratis för att spara det här jobbet.