Noam Elata

Generative AI Researcher at NVIDIA
Ph.D. Graduate at Technion — Israel Institute of Technology
Generative models, computer vision, and deep learning

About Me

I am a generative AI researcher at NVIDIA in Santa Clara, working in the Cosmos Lab on generative AI and world modeling. I work on generative models such as LLMs and diffusion models, studying their theoretical foundations, pre-training and scaling, neural architecture design, and efficient inference. I recently completed my Ph.D. in Electrical and Computer Engineering at the Technion, advised by Prof. Michael Elad and Prof. Tomer Michaeli. My doctoral research centered on image diffusion models, from theory and training algorithms to their use in inverse problems. I also had the opportunity to work with Prof. Daniel Soudry on attention sparsity and deep learning theory. Alongside my Ph.D., I was a computer vision researcher at Apple in Herzeliya, first part-time and then full-time, working on image and video diffusion models.

Prior to my Ph.D. studies, I received my B.Sc., summa cum laude, in Computer Engineering from the Technion. During my studies, I worked as a computer vision engineer at Mobileye.

Publications

2026

Noam Elata*, Itay Lamprecht*, Mikey Shechter*, Daniel Ohayon, Itay Hubara, Daniel Soudry
More Value per Key: Asymmetric Sparse Attention for Faster LLM Decoding
NeurIPS 2026 — The 40th Conference on Neural Information Processing Systems
code

Daniel Ohayon, Itay Lamprecht, Itay Hubara, Israel Cohen, Daniel Soudry, Noam Elata
Block Sparse Flash Attention
NeurIPS 2026 — The 40th Conference on Neural Information Processing Systems
code

Shai Yehezkel, Shahar Yadin, Noam Elata, Yaron Ostrovsky-Berman, Bahjat Kawar
Accelerating Text-to-Video Generation with Calibrated Sparse Attention
ECCV 2026 — The 19th European Conference on Computer Vision

2025

Noam Elata*, Hyungjin Chung*, Jong Chul Ye, Tomer Michaeli, Michael Elad
InvFusion: Bridging Supervised and Zero-shot Diffusion for Inverse Problems
NeurIPS 2025 — The 39th Conference on Neural Information Processing Systems
code

Noam Elata, Bahjat Kawar, Yaron Ostrovsky-Berman, Miriam Farber, Ron Sokolovsky
Novel View Synthesis with Pixel-Space Diffusion Models
CVPR 2025 — The IEEE/CVF Conference on Computer Vision and Pattern Recognition
code

Noam Elata, Tomer Michaeli, Michael Elad
PSC: Posterior Sampling-Based Compression
TMLR 2025 — Transactions on Machine Learning Research
code

2024

Shahar Yadin, Noam Elata, Tomer Michaeli
Classification Diffusion Models: Revitalizing Density Ratio Estimation
NeurIPS 2024 — The 38th Conference on Neural Information Processing Systems
webpage code

Noam Elata, Tomer Michaeli, Michael Elad
Adaptive Compressed Sensing with Diffusion-Based Posterior Sampling
ECCV 2024 — The 18th European Conference on Computer Vision
☆ Rothschild Academic Excellence Award
code

Bahjat Kawar*, Noam Elata*, Tomer Michaeli, Michael Elad
GSURE-Based Diffusion Model Training with Corrupted Data
TMLR 2024 — Transactions on Machine Learning Research
code

Noam Elata, Bahjat Kawar, Tomer Michaeli, Michael Elad
Nested Diffusion Processes for Anytime Image Generation
WACV 2024 — The IEEE/CVF Winter Conference on Applications of Computer Vision
webpage code space

Awards

NeurIPS Top Reviewer, 2025
Rothschild Academic Excellence Award, 2024
Meyer Fellows Prize, 2022
EMET Excellence Program, 2020–2022
Alfred and Anna Grey Excellence Scholarship, 2021
Apple Excellence Award, 2021
Technion Alumni Scholarship, 2020

Teaching & Program Committee

I have taught the following courses as a TA:

Generative AI — Diffusion Models (CS236610, CS236759) – Winter 2023–2024 (recorded lectures), Spring 2025
Deep Learning (ECE046211) – Spring 2023 (projects GitHub)

I have served as a reviewer at the following venues:

NeurIPS (2025 Top Reviewer, 2026)
ICML (2025, 2026)
CVPR (2025, 2026)
ECCV (2024, 2026)
ICLR (2025)