Building the future of Insurance.

CTO at OpenEyes. Formerly Samsung Exec. PhD in Image Synthesis through Weak Supervision.

Omry Sendik

Current

CTO & Co-Founder

OpenEyes • 2021 — Present

I'm the CTO and Co-Founder of OpenEyes, where we're reinventing Commercial Auto Insurance using Computer Vision, Generative AI, and by closing the loop on loss runs. We combine state-of-the-art tech with insurance models to help fleets drive safer and save more, because crashes are expensive, but preventable. Before taking the startup plunge, I spent over a decade at Samsung, most recently leading a brilliant team of engineers in the automotive algorithms realm. We tackled everything from Image Signal Processing to Machine Learning and Computer Vision. I hold a PhD in Computer Science from Tel Aviv University, where I explored how Generative Neural Networks can dream up images with no supervision. My MSc, on the border between Electrical Engineering and Applied Math, also from TAU, was all about sampling theory, and I started my journey at the Technion with a BSc in EE and a BA in Physics. Probably most importantly, and when I'm not knee-deep in pixels or insurance data, I'm a proud but often absent father of three and husband.

Research

Computer Vision, Generative Models & Signal Processing

ACM TOG 2020

uMM-GAN: Unsupervised K-Modal Styled Content Generation

A novel architecture building on StyleGAN to model multi-modal distributions in a completely unsupervised fashion.

O. Sendik, D. Lischinski, D. Cohen-Or

View Paper →
Image Communication 2019

DeepAge: Deep Learning of face-based age estimation

Dual CNN and SVR approach for age estimation using representation and metric learning.

O. Sendik, Y. Keller

View Paper →
WACV 2019

CrossNet: Latent Cross-Consistency for Unpaired Image Translation

Introducing latent cross-consistency to regularize unpaired image translation operators.

O. Sendik, D. Lischinski, D. Cohen-Or

View Paper →
CVPR 2019

IM-Net for High Resolution Video Frame Interpolation

Formulating motion estimation as classification to outperform previous methods on HD video datasets.

T. Peleg, P. Szekely, D. Sabo, O. Sendik

View Paper →
Eurographics 2019

What's in a Face? Metric Learning for Face Characterization

Determining characteristic facial parts using deep features and weakly supervised metric learning.

O. Sendik, D. Lischinski, D. Cohen-Or

View PDF →
ACM TOG 2017

Deep Correlations for Texture Synthesis

Introducing structural energy based on correlations among deep features to capture structural regularities.

O. Sendik, D. Cohen-Or

View PDF →
Atmospheric Research

Reconstructing the moisture field using microwave data

Reconstructing 2D humidity fields using cellular network opportunistic microwave links.

N. David, O. Sendik, et al.

View Paper →
IEEE Signal Processing Magazine

Precipitation Monitoring: A critical survey

A critical survey on multidimensional signal processing for weather monitoring via CWCNs.

O. Sendik, H. Messer

View Paper →
IEEE Trans. MTT

A 90-nm CMOS Power Amplifier for WiMAX

Single-stage CMOS power amplifier for 2.3-2.7-GHz WiMAX applications.

O. Degani, et al.

View Paper →

Academics & Theses

PhD Image synthesis through weak supervision View
MSc On the Coverage and Reconstructability of 2D Functions View
BSc Speech Enhancement Employing Discrete Modulation Transforms View
BSc Analytical Mechanics View