Mara Graziani

The Research Scientist

At the Panoramai AI Summit's closing session, Mara Graziani, Research Scientist at IBM Research Europe, provided an intimate look into cutting-edge AI research that bridges theoretical innovation with practical business transformation. Drawing on her decade-long journey in AI development, she illuminated how fundamental research translates into revolutionary applications across industries.

The IBM Research Innovation Framework

Graziani positioned IBM Research as the strategic incubator for next-generation AI solutions: « We really like to think of us as the incubators of new ideas, of innovation, of what's coming next. » She outlined their mission to empower organizations through generative AI tools, working synergistically across three core areas: developing new architectures that integrate multiple modalities, advancing synthetic data generation, and implementing generative workflow systems.

« We really want to empower everybody, all the people that work with us, all the collaborators that we have and the partnerships, to help them, to empower them with the generative AI tools, » she explained, emphasizing their focus on real-world applications with actual data and genuine business problems.

Revolutionary Digital Twins: Beyond Traditional Modeling

Graziani's most compelling breakthrough involved AI-powered digital twins that fundamentally transform predictive modeling. Describing a successful sensor lifecycle prediction project, she revealed: « The revolution we are seeing with digital twins that are powered by AI is that you can predict that and model it actually quite accurately. And the benefit of doing this is that you don't have to wait that time. »

The transformation represents a paradigm shift from physics-based to data-driven approaches. « Before we needed to know the physical equations of a system, we needed to understand very deeply the system to be able to model it, which is something that now we can bypass because we can have data-driven approaches, » she explained. This breakthrough enables modeling of previously inaccessible complex systems, from battery lifespan prediction to weather forecasting through NASA collaborations.

Synthetic Data Generation: Democratizing AI Creation

Graziani highlighted IBM's InstructLab project as a revolutionary approach to synthetic data generation that transforms users into « creators of value because they can use their own taxonomy and their own interpretation of the world and their own domain knowledge to build better models. »

The system creates domain-specific taxonomies, generates corresponding synthetic data, and fine-tunes language models at high throughput levels using quantized models that don't require massive infrastructure. « Data is where a lot of work needs to be done, right? A lot of companies have a lot of data, but it requires a lot of cleaning and preparation for all of the machine learning pipelines, » she observed, positioning synthetic data as a solution to bypass traditional data preparation bottlenecks.

The AI Evolution: From Manual to Autonomous

Reflecting on her ten-year research journey, Graziani traced AI's dramatic transformation. Starting in 2015, she experienced the field when « there was machine learning, but it was a lot of hand crafting of features, a lot of domain knowledge needed and it was obviously very expensive because you had to take all of those data sets and annotate them. »

The 2016-2018 revolution introduced self-supervision and foundation models that dramatically reduced costs and democratized AI development. This transformation enabled greater risk-taking and innovation: « You also increase the return of investment if you want because whereas before you had to put all of your investments in taking these data sets and labeling them and annotating them, now you can... focus only on having very small, very high quality data set that you can tune on. »

Strategic Research Focus in Accelerating Innovation

Despite rapid technological advancement, Graziani emphasized IBM Research's disciplined approach: « We still keep quite focused. It didn't make us lose the focus. » The team continues investing significant time in strategic thinking, using conferences and community engagement to gather feedback and understand field evolution.

She identified the integration of AI with physical world understanding as the next frontier: « That's really where you see the revolution happening, » noting that AI-powered digital twins can now model systems for which traditional physical understanding was previously impossible or prohibitively complex.

Data-Driven Discovery Acceleration

Graziani's vision extends to transforming scientific discovery itself. « When you have a lot of data, you can infer those from the data. And that's how you learn much faster and you accelerate the pace of discovery, » she concluded, positioning AI as a fundamental accelerator of human understanding across domains from materials science to healthcare.

Practical Innovation Philosophy

Her approach to making previously « unexploitable » data valuable through reinforcement learning and advanced modeling techniques demonstrates IBM Research's commitment to solving real business challenges while advancing fundamental AI capabilities.

Key Achievement: Graziani demonstrated how fundamental AI research can solve previously intractable real-world problems, positioning IBM Research as a bridge between cutting-edge innovation and practical business transformation while accelerating scientific discovery across industries through data-driven approaches that bypass traditional modeling limitations.

A computer scientist specializing in multimodal foundation models for scientific discovery at IBM Research Europe. She earned her PhD in Computer Science from the University of Geneva in 2021, where her research focused on developing methods to produce interpretable outputs from deep learning approaches without compromising performance. Her academic journey includes a visiting research position at the Martinos Center at Harvard Medical School, where she explored interactions between clinicians and deep learning systems, and a Master of Philosophy in Machine Learning, Speech and Language from the University of Cambridge. With undergraduate roots in Information and Communication Technology Engineering from Sapienza University of Rome, where she researched EMG signals for non-invasive hand prosthetics, she brings a multidisciplinary perspective to her work. Fluent in English, French, and Italian (native), she bridges international research communities in her pursuit of more interpretable and effective AI systems for scientific applications.