About the Lab
The Efficient Intelligence Lab at Technion–Israel Institute of Technology researches efficient reasoning and intelligence, aiming to close the gap between small and large reasoning models. The central question it asks is: What would AI look like if it learned and reasoned within the data and computing budget of a single human, not a global datacenter?
In recent years, progress in AI has mostly come from increasing computing power and data, running larger models, creating larger GPU clusters, and feeding them more text, code, and interaction data. Its costs are inaccessible to most researchers and institutions.
The Efficient Intelligence Lab focuses on the sufficiency frontier, aiming to develop principles that let smaller models learn more from less and retain much of the reasoning power of today’s largest systems, while operating within power, hardware, and data budgets that the real world can afford.
Research spans deep learning theory (including scaling laws and the expressive power of neural networks), AI-accelerated quantum many-body physics simulations, and large language models—both frontier-scale models in industry and more efficient, structured reasoning models in academia.
The lab connects foundational theory, scientific applications, and modern reasoning models into a unified focus on efficient intelligence.
Scholar Profile
At the Hebrew University of Jerusalem, Dr. Sharir’s doctorate in computer science tried to understand why certain neural network architectures are better than others in terms of the complexity of the tasks they can perform versus their overall computational budget.
His postdoctoral research at Caltech’s Department of Computing + Mathematical Sciences deepened his exploration of efficiency-driven approaches to machine learning.
In his lab at Technion–Israel Institute of Technology’s Faculty of Data and Decision Sciences , Dr. Sharir works on building a robust theoretical foundation for evaluating the strengths and weaknesses of newly proposed computing architectures.
