Ron Wein Email & Phone Number

New York, Us

Boston, Massachusetts, Us

I established the Speech Recognition team, responsible for developing a proprietary transcription engine and training the models used by the engine.The engine is developed in C++, where the main models are trained in Kaldi or PyTorch. Additional algorithmic frameworks include signal processing, HMM, WFST, deep models (including TDNN, RNN, CNN), graph algorithms, statistical processing.The engine is integrated in LogMeIn's VoIP products such as GoToMeeting and GoToWebinar and in its telephony products such as Grasshopper and GoToConnect (formerly Jive).Main accomplishments:• Developed a state-of-the-art ASR engine, superior to most other ASR vendors (according to internal benchmarks), including high-quality punctuation capabilities.• Optimized the CPU and memory consumption of the engine, so it can serve dozens of parallel sessions on a single server.• Development of numerous peripheral signal-processing algorithms, language model customizations, spoken language identification, and more.• Team building: Recruiting and training researchers and programmers in the Speech Recognition field.

Niskayuna, New York, Us

As a member of the AI Team in Hertzliya, I was involved in various projects for different departments in the company, such as data mining for transportation optimization (tabular data), deep-learning techniques for medical imaging, image segmentation for industrial inspection, etc.

Melville, New York, Us

I led a team of 5-8 researchers and linguists, responsible for the development of Automatic Speech Recognition modules for recorded call-center telephony sessions, and for the next generation of Speech Analytics algorithms applied on the transcribed sessions.The team was responsible for developing and supporting models for 20+ languages, and specialized in various algorithms such as language-model customization, speaker adaptation, diarization, classic and statistical NLP algorithms, sentiment analysis, etc.

Working in the "Madgera", I led a project ordered by Plurality Ltd.I developed a speech-recognition engine suitable for a highly parallel multi-core processor, developed by Plurality.

Il

RadLive developed a VoIP-based audio-conferencing system with enhanced quality and features.I was responsible for the development and implementation (on a TI 64+ DSP) of all the algorithmic modules, including:- Noise suppression,- Acoustic echo cancellation,- Audio compression,- Packet-loss concealment,- Speaker identification, etc.

Tel Aviv, Il

I worked in Prof. Dan Halperin's group in the field of Computational Geometry.My PhD thesis is titled "The Integration of Exact Arrangements with Effective Motion Planning" (approved Oct. 2007, with distinction).

Us

Until December 2002, I led a team that was responsible for the speech recognition algorithms and for developing training tools for hidden Markov models (HMMs).From January 2003 I worked as a part-time consultant.

Phd, Computer Science

Msc, Computer Science

Bsc, Mathmatics And Computer Science