Hi! I am a Research Scientist.
My work focuses on the design, analysis and discovery of novel algorithms for next-generation computing, particularly at the intersection of Artificial Intelligence and quantum technologies. My research portfolio includes foundational contributions to artificial intelligence, from the development of multimodal foundation models to their rigorous evaluation in terms of reliability, trustworthiness and uncertainty.
This website hosts a collection and summary of my work, talks and hints on the small habits that keep me living a happy and satisfying life.
Contact me>> Publications >> Selected Press >> Mentoring
WORK EXPERIENCE
Research Scientist → IBM Research Europe, Zurich, Switzerland, 2021–Now
Scientific Research Assistant → Hes-so Valais, Sierre, Switzerland, 2016–2023
SELECTED TALKS
IBM Think on Tour, Milano, IT, 2023 → Generative AI for Business
Royal Society of Medicine, London, UK, 2023 → “The human-AI horizon in microscopy imaging“
Banff International Research Station, Banff, Canada, 2022 → “DL Interpretability for the discovery of biomedical patterns”
Argonne National Laboratory, Chicago, USA, 2018 → “Regression Concept Vectors for Bidirectional Explanations in AI”
SELECTED PRESS
IBM Research Blog, 2025→ How to make AI models more accurate: Embrace failure.
L’invitée de La Matinale, 2023 → Mara Graziani, Chercheuse en IA
Le Temps 2022, Forum des 100 → Mara Graziani, Chercheuse spécialiste en intelligence artificielle
Le Temps 2022, Tête Chercheuse → Mara Graziani: «Je peux avoir une idée le matin et la tester tout de suite, le jour même»
SELECTED PUBLICATIONS
ICLR 2026 → Test-Time Tuned Language Models Enable End-to-end De Novo Molecular Structure Generation from MS/MS Spectra
Science Advances → Negative chemical data boosts language models in reaction outcome prediction
Transactions in Machine Learning Research → Uncovering Unique Concept Vectors through Latent Space Decomposition
Nature Communications → Regression-based Deep-Learning predicts molecular biomarkers from pathology slides
CVPR 2023 XAI4CV → Disentangling neuron representations with concept vectors
ICLR 2023 Trust (Un)-Limited →Concept discovery and dataset exploration with singular value decomposition