A Researcher in Multimodal Large Language Models
Pubblicato il 01-08-2026 - Altro in Italia
A Researcher in Multimodal Large Language Models (LLMs)
Join to apply for the A Researcher in Multimodal Large Language Models (LLMs) role at Fondazione Bruno Kessler - FBK
The Machine Translation (MT) Unit pursues research in the field of automatic translation, both from text and speech, to enable multilingual communication across modalities. Its core activities include the creation of advanced models and their adaptation to diverse contexts, domains, and applications (e.g. simultaneous speech translation and subtitling), leveraging the latest technology, including LLMs and foundation models. The unit also conducts fundamental and applied research that focuses on the ethical and societal impact of translation technology, with a long history in the design of evaluation methods, resources and studies that include human participants (e.g. gender inclusive and human-centered AI).
Job Description
The successful candidate is expected to research, design, and develop advanced models for multimodal (speech and video) processing, with a focus on novel architectures, fine‐tuning strategies,
integration of background knowledge (e.g., documents from knowledge bases), integration of speech and video modalities into LLMs, and optimization for computational efficiency. The candidate will also contribute to the evaluation and integration of these solutions in real-world use cases.
The successful candidate will contribute to the following tasks:
Conduct research in one or more of the following areas: multimodal (speech, video) integration, speech translation, and low-resource fine-tuning;
Design, develop, and deploy advanced machine learning algorithms, with a focus on efficiency;
Collaborate with team members to integrate research outcomes into practical solutions;
Contribute to publications and present research outcomes.
Job Requirements
The ideal candidate should have:
Master's Degree in areas related to Computer Science;
Solid background in deep learning;
Strong theoretical kno