Aviv Shamsian

PhD // Data Science Tech Lead at aiOla

Download PDF

Summary

Research Scientist with a PhD and first-author publications at leading ML conferences. Expertise in multi-task learning, learning in weight-space methods, and generative AI for speech. Skilled at translating research breakthroughs into large-scale production systems.

Education

PhD in Machine Learning (2022 – 2025) Faculty of Engineering, Bar Ilan University Thesis: Learning with Multiple Tasks Advisors: Prof. Gal Chechik, Prof. Ethan Fetaya

M.Sc in Electrical Engineering (2019 – 2021) Bar Ilan University Supervised by Prof. Gal Chechik

B.Sc in Medical Engineering (2013 – 2017) Afeka College of Engineering GPA: 92, Dean’s Honor

Experience

Data Science Tech Lead (2023 – Present) aiOla

  • Responsible for the technical design and architecture of data science projects, while also driving research through publications in top-tier speech conferences and innovation through patents to strengthen aiOla’s IP
  • Designed and implemented a proprietary ASR model with zero-shot contextual biasing to jargon and domain-specific terms, achieving 93% multilingual transcription recall and 30% reduction in WER for industrial use cases
  • Developed a proprietary keyword-spotting model with open-vocabulary zero-shot capabilities achieving >95% F1 on multilingual zero-shot recognition

Applied Scientist (2021 – 2023) Lab126, Amazon

  • Led the development of a fitness model for tracking and analyzing user exercises, providing personalized feedback
  • Drove the project from research and architecture exploration through experimentation, implementation, and deployment on mobile edge devices

Chief Technical Advisor (2019 – 2021) DeepWrex

  • Led a research team on cutting-edge deep learning research in the image restoration domain

Data Scientist (2019 – 2020) DragonTail Systems

  • Developed novel Machine learning and Deep learning approaches for recommendation systems and computer vision applications
  • Led the end-to-end project of segmenting different chicken pieces for KFC, resulting in lower food costs and improved customer service

Machine Learning Researcher (2016 – 2019) spirCare Ltd

  • Developed machine learning-based models for estimation and regression problems
  • Algorithm developer in Matlab & Python environment - physiological signal processing

Selected First-Author Publications

  • Keyword-guided Adaptation of Automatic Speech Recognition, InterSpeech 2024
  • Improved Generalization of Weight Space Networks via Augmentations, ICML 2024
  • Auxiliary Learning as an Asymmetric Bargaining Game, ICML 2023
  • Personalized Federated Learning using Hypernetworks, ICML 2021
  • Learning Object Permanence from Video, ECCV 2020

Full publication list

Volunteering

“Big Brother” Project (2018 – 2023) As part of the Big Brother project, we accompany youth at risk and guide them