About
Nov. 2023-Present: Research scientist at the Applied machine learning lab, Juelich Supercomputing Center.
Feb. 2022-Oct. 2024: Course instructor for undergrad CS courses at the Open University of Israel (AI, Computer Graphics, Data Structures and Algorithms).
Dec 2017-Aug 2019, Feb. 2021-Aug. 2022, Aug. 2022-Nov. 2023. Independent PhD candidate in computer science, computer vision lab, University of Bonn.
Additional responsibilities included teaching and supvervising Master’s student, e.g. Adversarial Synthesis of Human Pose from Text, which won the best Master’s thesis award in DAGM Young Researchers’ Forum 2020.
Due to a personal loss, I experienced delays in my research in 2023-2024.
Research interests
During my PhD, I conducted research on
- Human Pose estimation and dabbled with ordinal prediction for sets in the process
- Weakly supervised semantic segmentation using object size constraints.
- Optical flow and scene flow for human body tracking during my internship at Facebook Reality Labs
- Generating human poses conditioned on text
- Generating human motion conditioned on multiple actions while interning at Amazon, followed up by scene constrained generation.
MSc in computer science, University of Bonn.
BSc in computer and software engineering, Technion.
Research internships:
Applied scientist intern at Amazon Go, August 2021-August 2022. My work focused on synthesizing conditional human motion data.
Research intern at Facebook Reality Labs, August 2019-Jan 2020. I worked on differentiable rendering and optical flow for human body tracking.
Master’s thesis
Online Robust Learning Using the Radon (center) Point. I proposed a method for aggregating multiple models that are trained independently using multiple data streams in a federated learning setting. Earlier works rely on simply averaging the models using the arithmetic mean, which fails when the data is noisy. The proposed method achieved superior error bounds both in theory and practice in the presence of outliers in the training data.