One Postdoctoral position multimodal and generative AI physical and natural sciences

Position: One Postdoctoral position on multimodal and generative AI for physical and natural sciences
Organisation/Company Istituto Italiano di Tecnologia Research Field Computer science Researcher Profile Recognised Researcher (R2) Final date to receive applications 31 May 2026 - 00:00 (UTC) Country Italy Type of Contract Other Job Status Other Is the job funded through the EU Research Framework Programme? Not funded by a EU programme Is the Job related to staff position within a Research Infrastructure?

No

Offer Description Commitment & contract: 12 months (+12 renewable)

Location:

Genova Step into a world of endless possibilities, together let’s leave something for the future!

At IIT, we are committed to advancing human-centered Science and Technology to address the most urgent societal challenges of our era. We foster excellence in both fundamental and applied research, spanning fields such as neuroscience and cognition, humanoid technologies and robotics, artificial intelligence, nanotechnology, and material sciences, offering a truly interdisciplinary scientific experience. Our approach integrates cutting‑edge tools and technology, empowering researchers to push the limits of knowledge and innovation.

With us, your curiosity will know no bounds.

We are dedicated to providing equal employment opportunities and fostering diversity in all its forms, creating an inclusive environment. We value the unique experiences, knowledge, backgrounds, cultures, and perspectives of our people. By embracing diversity, we believe science can achieve its fullest potential.

THE ROLE You will be working in a multicultural and multi‑disciplinary group, where junior and senior scientists collaborate, each with their expertise, to carry out a scientific activity with shared research goals. The research focuses on fundamental AI topics from methodological and theoretical perspectives, yet functional to tackle a number of applications and actual case studies related to several domains.

The Artificial Intelligence for Good (AIGO) research unit is coordinated and led by prof. Vittorio Murino.

AIGO benefits from the collaboration with several universities and research centres worldwide, most often with the closer universities of Genova and Verona. AIGO is part of ELLIS – an European network of excellence in AI, Machine Learning (ML) and Computer Vision (CV), of which Vittorio is a Fellow member.

For this particular position, the focus is on investigating AI approaches for physical and natural sciences, e.g., physics, chemistry, material science, weather forecast, biology, neuroscience, etc. Interestingly, these domains are characterized by the availability of huge amount of multimodal data (e.g., meteorological acquisitions, with radar scans and other ambient or physical parameters), and this makes AI techniques particularly convenient to use, since they are particularly data‑hungry.

But many lines of research in these domains have been faced to date by designing particular simulation systems, mainly based on physical laws (e.g., differential equations). To this end, a possibility to be explored is the design of hybrid simulation systems in which the rigor of physical laws is integrated with the power and versatility of AI methods trained with on‑field data, either annotated or not.

In this context, the fast development of always novel generative models (from GANs to Flow Matching) will open other possibilities in the study of physical sciences’ applications. Finally, the data‑agnostic nature of these models makes AI particularly suitable to cope with these issues, being possible to deal with multimodal data such as images, signals, strings, but also text.

AIGO is a perfect environment to study these topics given its expertise in Machine and Deep Learning, Computer Vision, Signal Processing, and Multimedia. Also, its declared vision to work especially in presence of imperfect data and multimodal setups – hence by tackling unsupervised, semi‑supervised and self‑supervised settings, weakly or noisy labelled, few, class imbalanced, or biased data – is exactly what it is needed to cope with these scenarios.

Domain adaptation, generalization, few/zero‑shot learning, and open‑set recognition, topics in which AIGO has…