Ben Fishman

AI Researcher

I am a Ph.D candidate in the Computer Science department at Bar Ilan University, supervised by Prof. Gal Chechik and Dr. Idan Schwartz. My research focuses on understanding Deep Neural Networks, with a particular emphasis on Multi-Modality and Disentanglement in Generative Networks.

Prior to my Ph.D studies, I served as the Director of AI & Algorithms at Microsoft, leading the Data Science discipline in the Surface and WSSI groups. My team worked on Computer Vision, securing LLMs and DNNs compression. Before that, I was a Senior Data Scientist in Microsoft’s Video Indexer group, focusing on developing Deep Learning models for audio and video modalities in the media and autonomous driving domains. I also worked at Mobileye as an Algorithms Developer and Team Leader, specializing in Computer Vision, Deep Learning, Signal Fusion, and Temporal Analysis.

I hold an M.Sc. in Electrical Engineering from Tel Aviv University, where I researched Speech, specifically Prosody (intonation, rhythm, timbre, etc.). I also have a B.Sc. in Biomedical Engineering from Tel Aviv University, where my research centered on detecting human gait and movement using Sonar.

I am passionate about understanding images, videos, music, and speech.

news

Apr 24, 2025	My lecture from IMVC 2025, “Bringing AI to Production,” is now live. You can now watch the full video and check out the slide deck from the talk. Click here
Feb 15, 2025	My talk “Bringing AI to Production” was accepted to IMVC. See you there - April 1, 2025.
Feb 01, 2025	My webinar “Evaluation of AI models and Algorithms” was accepted to the IMVC pre conferance webinar series. See you on - March 11, 2025.
Jan 01, 2025	I started my Ph.D studies! More to come, soon…

selected publications

Paper

Prosodic Feature Criterion for Hebrew Using Different Feature Sets

Ben Fishman, and Irit Opher

In IEEE International Conference on the Science of Electrical Engineering in Israel (ICSEE), Sep 2018

Bib HTML PDF

@inproceedings{fishman2018prosodic,
  title = {Prosodic Feature Criterion for Hebrew Using Different Feature Sets},
  author = {Fishman, Ben and Opher, Irit},
  booktitle = {IEEE International Conference on the Science of Electrical Engineering in Israel (ICSEE)},
  pages = {1--5},
  year = {2018},
  month = sep,
  organization = {IEEE},
}

Paper

Prosodic Features’ Criterion for Hebrew

Ben Fishman, Itshak Lapidot, and Irit Opher

In Springer International Conference on Text, Speech, and Dialogue, Dec 2018

Bib HTML PDF

@inproceedings{fishman2018prosodid,
  title = {Prosodic Features’ Criterion for Hebrew},
  author = {Fishman, Ben and Lapidot, Itshak and Opher, Irit},
  booktitle = {Springer International Conference on Text, Speech, and Dialogue},
  pages = {482--491},
  year = {2018},
  month = dec,
  organization = {Springer},
}

Paper

Noncontact Wideband Sonar for Human Activity Detection and Classification

Gaddi Blumrosen, Ben Fishman, and Yossi Yovel

IEEE Sensors Journal, Dec 2014

Bib HTML PDF

@article{blumrosen2014noncontact,
  title = {Noncontact Wideband Sonar for Human Activity Detection and Classification},
  author = {Blumrosen, Gaddi and Fishman, Ben and Yovel, Yossi},
  journal = {IEEE Sensors Journal},
  volume = {14},
  number = {11},
  pages = {4043--4054},
  year = {2014},
  publisher = {IEEE},
}

Paper

SLIP: Securing Llms IP using weights decomposition

Yehonathan Refael, Adam Hakim, Lev Greenberg, Tal Aviv, Satya Lokam, Ben Fishman, and Shachar Seidman

arXiv preprint, Dec 2024

arXiv Bib PDF

@article{refael2024slip,
  title = {SLIP: Securing Llms IP using weights decomposition},
  author = {Refael, Yehonathan and Hakim, Adam and Greenberg, Lev and Aviv, Tal and Lokam, Satya and Fishman, Ben and Seidman, Shachar},
  journal = {arXiv preprint},
  year = {2024},
}

Thesis - Prosodic Feature Criterion

Ben Fishman

Feb 2020

Bib PDF