Go directly to page contentent
TEAM LADE

Alessio Ansuini

Researcher at Laboratory of Data Engineering

I am a theoretical physicist with interdisciplinary experience across biology, physics, and computation. My research has addressed individual biological molecules and neurons to learning, cognition, and artificial intelligence. From 2011 to 2019, I collaborated with experimental and theoretical groups at SISSA, tackling problems in neurobiology, cognitive neuroscience, physics, and machine learning, and in 2019, I joined Area Science Park where I helped shaping its scientific research directions. My current focus involves understanding neural networks representations and exploring new frameworks for information processing in biological systems. Since 2018, I have been teaching Deep Learning courses at the University of Trieste and for the joint SISSA–ICTP Master’s in High-Performance Computing. In addition, I have delivered lectures at international schools and remain committed to public speaking and artistic initiatives aimed at sharing scientific ideas with a broad audience.

 

Research Interests 
  • Analysis of neural representations
  • Computational neuroscience
  • Artificial intelligence
Experience & Education
  • Master in High-Performance Computing, SISSA/ICTP, 2016
  • PhD in Theoretical Physics, “Sapienza” University of Rome, 2010
  • MSc in Physics, “Sapienza” University of Rome, 2005
Latest pubblications
21/09/2023
The geometry of hidden representations of large transformer models
Abstract Large transformers are powerful architectures used for self-supervised data analysis across various data types,…
Go to the news The geometry of hidden representations of large transformer models
16/07/2024
Enhancing predictions of protein stability changes induced by single mutations using MSA-based language models
Abstract Protein language models offer a new perspective for addressing challenges in structural biology, while…
Go to the news Enhancing predictions of protein stability changes induced by single mutations using MSA-based language models