Theme Adjuster

Adjust the sliders and copy the values below into your CSS to change the theme.

--primary-h: 216; --primary-s: 35%; --primary-l: 67%;

Efficient Intelligence Lab

Technion – Faculty of Data and Decision Sciences

Research on efficient reasoning and intelligence: closing the gap between small and large reasoning models.

Read Our Vision ↓ Prospective Students

Vision

Our work centers on a fundamental question:

What are the minimal computational and data requirements needed to support robust reasoning and intelligence?

Today’s AI progress is driven by scaling: ever-larger models trained and served on enormous compute budgets. While effective, the costs are inaccessible to most. The strongest open-source large reasoning models—such as Kimi and Qwen—approach the quality of proprietary systems like GPT-5 and Claude, yet require multi-GPU servers. Meanwhile, small reasoning models (SRMs) remain far behind.

The lab aims to close this gap. We develop new principles and mechanisms that enable efficient reasoning: models that achieve strong capabilities with dramatically reduced compute and data. The human brain runs on tens of watts, whereas today’s leading models require kilowatts. That gap isn’t just striking — it’s an invitation to rethink what efficient intelligence could look like.

About the Lab

Or Sharir

Or Sharir

Assistant Professor, Technion

The Efficient Intelligence Lab is led by Or Sharir, a recently appointed Assistant Professor in the Technion Faculty of Data and Decision Sciences. Or has worked in both academia and industry, most recently as a co-founder of AAI.

He obtained his Ph.D. in Computer Science from the Hebrew University of Jerusalem under the supervision of Prof. Amnon Shashua, and completed a postdoc at the California Institute of Technology (Caltech), hosted by Prof. Anima Anandkumar and Prof. Garnet Chan.

Visit sharir.org →

Research Focus

His research spans deep learning theory (including scaling laws and the expressive power of neural networks), AI for quantum many-body physics, and large language models—both frontier-scale models in industry and more efficient, structured reasoning models in academia.

The lab continues this trajectory, focusing on understanding and engineering intelligence under real computational constraints.

Opportunities for Students

We welcome exceptional MSc and PhD students who want to work at the intersection of theoretical insight and experimental system-building.

Who We Are Looking For

You are likely a good fit if you have:

  • Strong foundations in machine learning, mathematics, and algorithms
  • Solid engineering skills in Python and C++
  • Curiosity, rigor, and comfort working across both theory and practice

Admissions & Eligibility

The lab is based in the Technion Faculty of Data and Decision Sciences. Joining the lab requires admission to a Technion MSc/PhD track, typically within this faculty.

I’m primarily looking for students who will join the MSc/PhD programs in Data and Decision Sciences, though cross-faculty co-advising may be possible in select cases. If you are not currently enrolled but plan to apply to these programs and are interested in the lab, you are very welcome to reach out.

How to Join

If you are interested in joining the lab, please email or.sharir@technion.ac.il with:

  • Your CV
  • Your transcript
  • A short paragraph explaining why you want to join the lab and why you believe you are a strong fit

Email Application Materials

Or contact or.sharir@technion.ac.il with questions.