Osher Azulay

I’m a Fulbright postdoctoral researcher at the University of Michigan, working with Prof. Stella Yu. I work at the intersection of robotics, computer vision, tactile sensing, and machine learning, aiming to advance humanoid intelligence.

Previously, I earned my Ph.D. from Tel Aviv University in 2024, under the supervision of Dr. Avishai Sintov. My work focused on robotic in-hand manipulation, developing methods that leverage multimodal cues to enable more adaptive interaction.

Always happy to connect—feel free to reach out.

Email  /  CV  /  Scholar  /  LinkedIn  /  Github

profile photo

News

  • July 2025 — Started my postdoc at the University of Michigan.
  • Winter 2025 — Visiting Scholar at UC Berkeley’s AUTOLab.
  • April 2025 — Gave a talk at Columbia University’s ROAM Lab.
  • Dec 2024 — Invited talk at Bar-Ilan University, Computer Science Department.
  • Dec 2024 — Invited talk at the Technion, Mechanical Engineering Robotics Colloquium.
  • Nov 2024 — Received the Fulbright Postdoctoral Fellowship.
  • Oct 2024 — Defended my Ph.D. at Tel Aviv University.
  • Summer 2023 — Visiting Graduate Researcher at Rutgers University, Robot Learning Lab.
  • Summer 2022 — Robotics Intern Engineer at Unlimited Robotics.
  • 2023 — Received Honorable Mention for Excellence in Teaching at Tel Aviv University.
  • 2023 — Awarded the KLA Ph.D. Excellence Scholarship.
  • 2022 — Awarded the Prof. Nehemia Levtzion Scholarship for Outstanding Doctoral Students.

Selected Publications:
VIGOR VIGOR: Visual Goal-In-Context Inference for Unified Humanoid Fall Safety
Osher Azulay, Zhengjie Xu, Andrew Scheffer, and Stella X. Yu.
Under review.
project page

Unified fall mitigation + stand-up recovery distilled into an egocentric-depth policy.
vt Embodiment-Agnostic Navigation Policy Trained with Visual Demonstrations
Nimrod Curtis*, Osher Azulay* and Avishai Sintov.
Under review.
paper / code / video

We proposed ViDEN, a framework using visual demonstrations for scalable, collision-free navigation.

vt Visuotactile-Based Learning for Insertion with Compliant Hands
Osher Azulay, Dhruv Metha Ramesh, Nimrod Curtis and Avishai Sintov.
IEEE RA-L & IROS, 2025.
website / paper / code

Sim2real learning of robust precision insertion polices with compliant hands.
allsight AllSight: A Low-Cost and High-Resolution Round Tactile Sensor with Zero-Shot Learning Capability
Osher Azulay, Nimrod Curtis, Rotem Sokolovsky, Guy Levitski, Daniel Slomovik, Guy Lilling and Avishai Sintov.
IEEE RA-L & ICRA, 2024.
paper / video / code

Introducing AllSight, an optical tactile sensor with a round 3D structure designed for robotic inhand manipulation tasks
ag Augmenting Tactile Simulators with Real-like and Zero-Shot Capabilities
Osher Azulay*, Alon Mizrahi*, Nimrod Curtis* and Avishai Sintov.
ICRA 2024.
paper / code

Tackling the sim-to-real problem for high resolution 3D round sensors using bi-directional Generative Adversarial Networks.
hapticrl Haptic-Based and SE(3)-Aware Object Insertion Using Compliant Hands
Osher Azulay, Max Monastirsky and Avishai Sintov.
IEEE RA-L & ICRA, 2023.
paper / video

Exploring complaint hands characteristics for object insertion using haptic-based residual RL.
throw Learning to Throw With a Handful of Samples Using Decision Transformers
Max Monastirsky, Osher Azulay and Avishai Sintov.
IEEE RA-L & IROS, 2023.
paper / video

Exploring the use of Decision Transformers for throwing and their ability for sim2real policy transfer.
hapticmpc Learning Haptic-based Object Pose Estimation for In-hand Manipulation Control with Underactuated Robotic Hands
Osher Azulay, Inbar Meir and Avishai Sintov.
IEEE Transactions on Haptics, 2022.
paper / video / code

In-hand object pose estimation and manipulation using Model Predictive Control.
ops Open-Sourcing Generative Models for Data-driven Robot Simulations
Eran Bamani, Osher Azulay, Anton Gurevich, and Avishai Sintov.
Data-Centric AI workshop, NeurIPS, 2021
paper / oral

Exploring the possibility of investing the recorded data in a generative model rather than directly to a regression model for real-robot applications.
komodo Wheel Loader Scooping Controller Using Deep Reinforcement Learning
Osher Azulay and Amir Shapiro.
IEEE Access, 2021
paper / video / code

A deep reinforcement learning-based controller for an unmanned ground vehicle with a custom-built scooping mechanism.

Template from source code.