doi.bio/ori_kabeli
Ori Kabeli
Overview
Ori Kabeli is an Artificial Intelligence and Machine Learning researcher and entrepreneur. He is currently working on his second startup inside Google's Area 120 incubator. Kabeli has previously worked as an AI researcher at Meta and Facebook AI Research (FAIR). He co-founded SlickLogin, which was acquired by Google in 2014, and Chatbase, developed under Google's Area120 incubator.
Education
Kabeli studied at Tel Aviv University, Israel, where he obtained a Master of Science in Computer Science and a Bachelor of Science in Computer Science and Business Management, graduating Summa Cum Laude.
Career
Kabeli has worked on various projects throughout his career, including leading cutting-edge AI research initiatives at FAIR, such as esmatlas.com, Speech Conditioning, and more. He has also worked with the Google Research & Google Identity teams and GV (Google Ventures).
From October 2009 to 2012, Kabeli worked as a Software Team Leader, leading a team of 10 programmers and QA testers specializing in applications for Real Time Embedded environments. Prior to that, from September 2006 to September 2009, he worked as a Real-Time Embedded Software Developer, designing key modules for high-speed communication devices in Real-Time embedded environments.
Research
Kabeli has authored and contributed to several research papers in the fields of AI, Machine Learning, and Computer Science. Some of his notable works include:
- "Decoding speech perception from non-invasive brain recordings," published in Nature Machine Intelligence. This work focuses on decoding speech from brain activity using non-invasive methods, eliminating the need for risky brain surgery.
- "Evolutionary-scale prediction of atomic-level protein structure with a language model," published in Science. This research demonstrates the use of a large language model for direct inference of atomic-level protein structure, resulting in significantly faster and more accurate predictions.
- "Language models generalize beyond natural proteins," where Kabeli and his colleagues demonstrate that language models, even when only trained on sequences, can learn a deep grammar that enables the design of protein structures beyond natural proteins.
- "Online Self-Attentive Gated RNNs for Real-Time Speaker Separation," in which Kabeli and his co-authors evaluate the performance of a causal and real-time model converted from a non-causal state-of-the-art separation model in both online and offline settings.
- "High Fidelity Speech Regeneration with Application to Speech Enhancement," where a wav-to-wav generative model for speech enhancement is proposed, aiming to improve speech intelligibility.
Recognition
Kabeli has received several recommendations and endorsements on LinkedIn from colleagues and team members, highlighting his proficiency, team spirit, management skills, and contribution to the field of AI research and development.