• Inventor of lossless audio codec method used in Apple Lossless Codec and FLAC, used to create the largest collection of lossless audio at the time. (US patents 6043763, 6121904, 6125348, 6028541, 6118392)
• Inventor of Vector Quantized Autoencoder Convolutional Neural Network (see PhD thesis)
• Created virtual hearing aid software with support of Vint Cerf for Google
• Startup co-founder 4x
• Consultant / advisor to 15+ startups
• Deep technical expertise in web development, fullstack development, audio/video processing, networking, embedded development, Linux
• Stanford PhD in AI / Neural Networks. Advisee of David Rumelhart (backpropagation) and Robert Gray

2023 - 2024 - Apple - Software Engineer, Consulting

2020 - present - ContentGroove - Co-founder and builder of AI video startup funded by Andrew Ng's AI Fund. Development using Large Language Models LLM, Tensorflow, PyTorch, React, Python, OpenCV, ffmpeg, node.js, Ruby on Rails, AWS Lambda, S3, DynamoDB, Cloudformation.

2000 - present - Zillabit LLC - Principal of consulting firm.

2018 - 2019 - Udacity - Development using React, GraphQL, Python, node.js, Docker, Go, Postgres/SQL, Kafka, Airflow, and other technologies.

2017 - present - Finishsign - Founder. Created API-driven electronic signature service using React, Python, node.js, Java, AWS Lambda, S3, DynamoDB, Cloudformation.

2016 - 2018 - Phase 2 Industries - Development using React, React Native, iOS, Android, Java/J2EE, Postgres/SQL, and other web technologies.

2014 - 2016 - CrunchBase - Consulting engineer. Development on webapp using Ruby on Rails, Postgres, jQuery, HTML5. Architected move of data and webapp to Postgres from neo4j graph database.

2013 - Tilera - Consulting engineer. Audio codec development for Tilera multicore architecture.

2013 - Polaris Wireless - Consulting engineer. Development of Android app and MySQL-based server.

2012 - 2013 - Bazaarvoice - Consulting engineer. Development on webapp using J2EE/Spring, Hibernate, Postgres, jQuery, HTML5.

2011 - 2012 - Private client - Designed and built web application for client's internal use. Used Ruby on Rails, jQuery/Javascript, MySQL, constraint programming (Gecode), custom native Ruby gems, Node.js, RabbitMQ (AMQP), and Hudson/Jenkins (continuous integration). Deployed on AWS EC2.

2009 - 2011 - Jawbone (Aliph) - Consulting engineer. Developed C/C++ and Java software for communication, control, and updating of Jawbone hardware devices via Bluetooth and USB, on Android, RIM, iOS, Mac, Linux, and Windows. Used libusb, wxWidgets, libcurl, mongoose, NSIS (Windows installer), PackageMaker (Mac installer). Created device firmware for Jawbone Era product flash storage feature, using CSR SDK. Built and used test mechanisms with device for both end-user and factory software. Set up continuous build and integration system (buildbot/Python) for both software and device firmware.

2003 - 2008 - Neurotone, Redwood City, California - Co-Founder and Vice President of Engineering. Technology solutions for people with hearing loss. Created Mac/Windows/Linux client software using ffmpeg, wxWidgets, libcurl, NSIS (Windows installer). Client provided cross-platform audio and video playback with custom DSP and synchronized variable speed. With Vint Cerf's support, designed and created virtual hearing aid software for Google's use by their employees. Created server (Tomcat, J2EE/Struts and MySQL) for synchronizing client data, and allowing access by a care provider in a HIPAA compliant manner. Build/bring-up for embedded Linux system for ARM (Xscale) SOC based device with display and custom serial input hardware, using buildroot and uboot bootloader.

2003 - 2010 - Back Office Music - Co-Founder and Technology Manager. Built and deployed system for custom branded, digital media download stores for Grammy award winning bands, using Tomcat, J2EE/Struts, and MySQL.

2004 - MusicGremlin - Technical advisor, consulting. Provided key advice to the founders at an early stage on technical strategy for their portable music device and back-end service. MusicGremlin was acquired by SanDisk in 2008.

2003 - 2004 - Mediabolic - Senior engineer, consulting. Developed various software components for embedded Linux multimedia set-top boxes: audio and DSP, ALSA and audio device driver, CD ripping, custom serial device interface, personal audio recorder application, interfacing with external satellite tuner. Mediabolic was acquired by Rovi (formerly Macromedia) in 2006.

2004 - 8x8, Inc. - Technical advisor, consulting. Identified and evaluated potential solutions for SIP soft-client for this VOIP company, following which the company executed based on my recommendation. Identified and evaluated possible wireless SIP handset solutions.

2003 - Rentfair - Technical lead. Co-developed prototype web service for movie rental by mail using Tomcat, J2EE/Struts, Torque (ORM forerunner to Hibernate etc.) and MySQL.

2002 - 2003 - PieStorm, Palo Alto, California - Co-Founder with Professor Garret Okamoto of company developing advanced technology for 802.11. Development of 802.11 baseband PHY (including OFDM) in C model and MATLAB (via Mex).

2002 - Telehealth Broadband, Austin, Texas - Consulting engineer developing applications for embedded Linux platform, including H.323 recording application and live HTTP streaming for audio.

2000 - 2001 - Sony / Unsurface, San Francisco, California - Lead Software Engineer - consulting. Architecting, managing, and implementing key software technologies for startup in online media space. Providing strategic and technical direction. Key technologies used in implementations included signal processing and signal detection algorithms, Java, Java Security, HTTP, XML (DOM and SAX parsers), JDBC/SQL, C++/COM/ActiveX/ATL, Windows Media SDKs (DRM, Format, Encoding, Device Manager), Resin, IIS/ASP, VBScript, Perl, Install Shield, Wininet, various Win32 interfaces, ASPI, Cygwin, and Javascript.

2001 - Trinity Signal, Menlo Park, California - Senior Technology Consultant. Providing strategic and technical direction of engineering effort for early stage startup in wireless space.

2000 - Ikimas, Campbell, California - Chief Technical Officer, Acting. Providing strategic and technical direction and management of engineering team for early stage startup.

2000 - Myplay, Redwood City, California - Consulting engineer. Integrating Microsoft Digital Rights Management into the myplay "locker" online music storage system using Java and JBDC.

1997-2000 - Liquid Audio, Redwood City, California - Chief Scientist and Manager of Authoring Tools and DSP Group. Architect of company's audio and signal processing technology. Implemented, optimized, and/or integrated all the signal processing components in Liquid Audio's system: audio codecs (MP3, AAC, ATRAC3, AC-3, Lossless), audio watermarking, equalization, dynamic compression, sample rate conversion, etc. Inventor of majority of company's intellectual property, with 11 patents filed. In management role, led engineering and testing teams that developed industry-leading authoring tools (encoding) products for Windows, Macintosh, and command-line (DOS and Unix) platforms. Participated in entire life-cycle of development of three major releases of the system of server, client, and authoring tools. Frequent mentor and advisor to entire engineering team. Software work done using C++ / object-oriented design in cross-platform development environment (Microsoft Developer Studio, Unix, Macintosh). Architect of build procedures and source code control administrator and guru. Technical evangelism: lectured on AAC audio codec (at Audio Engineering Society) and watermarking (at UC Berkeley). Participation in MPEG standards committee.

1997 - Vxtreme, Inc., Sunnyvale, California - Senior Engineer, Compression Group. Research and development for video and audio compression components of internet streaming media technology. Thorough knowledge of video and audio compression systems (H.263, HVQ, CELP, G.723.1, TwinVQ) as well as general signal processing methodologies (resampling, color conversion, preprocessing) and real-time streaming over the internet. Involvement in the MPEG standards committee. Software implementation, integration, and optimization using C++ / object-oriented design in cross-platform development environment (Microsoft Developer Studio and Unix). In August 1997 Vxtreme was acquired by Microsoft and integrated into its Windows Media (then called Netshow) division in Redmond, WA.

1994-1995 - Quantum Corporation, Milpitas, California - Digital Design Consultant - Digital designer for hard drive PRML read channel ASIC. Designed using Verilog HDL and Synopsys. Sole designer for key system components.

1991-1995 - Stanford University - Research Assistant to Professor David Rumelhart - investigating new neural network structures and applications.

1992 - Ricoh Research and Development Center, Yokohama, Japan - Researcher - neural network applications for optical character recognition.

1991 (Summer) - Ricoh California Research Center, Menlo Park, California - Researcher - project on neural network lipreading system for improved speech recognition.

1989-1990 - Stanford University - Teaching Assistant - classes in EE department: Fourier Transforms and Analog and Digital Laboratories.

1989 (Summer) - National Instruments, Austin, Texas - Engineer - developed debugger (including assembler and disassembler) for TI DSP product for Macintosh computer.

Education:

Ph.D., Electrical Engineering - Stanford University - 1996. GPA 3.96.

Research Supervisor: Prof. Robert Gray - Signal Compression and Classification Group

Studied: video compression, audio compression, speech compression, standards (MPEG, MPEG2, JPEG, H.261, H.324, H.263, G.723, Dolby AC-3, CELP, etc.), VQ, lossless coding (Huffman and Arithmetic), digital communications, digital design for ASIC (Verilog HDL, Synopsys), computer architecture, RISC, DSP programming, C++, C, UNIX. With David Rumelhart studied neural networks.

Dissertation: Stochastic Vector Quantization Using Neural Networks.

MSEE - Stanford University - 1992.

BSEE with High Honors - University of Texas at Austin - 1989.

Publications and Patents:

Kearby, G., Levine, E. & Modeste, R. Personal hearing suite. US Patent Number 8010366, issued August 30, 2011.

Brownell, J., Ansell, S., Devine, P. & Levine, E. Security mechanism for computer processing modules. US Patent Number 7549173, issued June 16, 2009.

Brownell, J., Ansell, S., Devine, P. & Levine, E. Security mechanism for computer processing modules. US Patent Number 7225330, issued May 29, 2007.

Levine, E. & Wiser, P. Server-side watermark data writing method and apparatus for digital signals. US Patent Number 7185200, issued February 27, 2007.

Brownell, J., Ansell, S., Devine, P. & Levine, E. Security mechanism for computer processing modules. US Patent Number 6965994, issued November 15, 2005.

Levine, E. Robust watermark method and apparatus for digital signals. US Patent Number 6345100, issued February 5, 2002.

Levine, E. Determination of a best offset to detect an embedded pattern. US Patent Number 6330673, issued December 11, 2001.

Levine, E. Secure watermark method and apparatus for digital signals. US Patent Number 6320965, issued November 20, 2001.

Levine, E. & Brownell, J. Robust watermark method and apparatus for digital signals. US Patent Number 6209094, issued March 27, 2001.

Levine, E. Efficient watermark method and apparatus for digital signals US Patent Number 6219634, issued April 17, 2001.

Levine, E. Lossless data compression with low complexity. US Patent Number 6125348, issued September 26, 2000.

Levine, E. Lossless data compression with low complexity. US Patent Number 6121904, issued September 19, 2000.

Levine, E. Lossless data compression with low complexity. US Patent Number 6118392, issued September 12, 2000.

Levine, E. Lossless data compression with low complexity. US Patent Number 6043763, issued March 28, 2000.

Levine, E. Lossless data compression with low complexity. US Patent Number 6028541, issued February 22, 2000.

Levine, E. & Chou, P. Error correction and loss recovery of packets over a computer network. US Patent Number 6000053, issued December 7, 1999.

Levine, E. Stochastic Vector Quantization, and Stochastic VQ With State Feedback Using Neural Networks. Proceedings of the 1996 IEEE Data Compression Conference, Snowbird, UT. Winner of the Renato Capocelli Best Student Paper Award.

Levine, E. A Time-Warping Neural Network. Proceedings of the 1995 International Conference on Acoustics, Speech, and Signal Processing, Detroit, MI.

Stork, D. G., Wolff, G., & Levine, E. Neural network lipreading system for improved speech recognition. Proceedings of the 1992 International Joint Conference on Neural Networks, Baltimore, MD.

Stork, D. G., Wolff, G., & Levine, E. Method and apparatus for extracting speech related facial features for use in speech recognition systems. US Patent Number 5771306, issued June 23, 1998.

Stork, D. G., Wolff, G., & Levine, E. Neural network acoustic and visual speech recognition system. US Patent Number 5586215, issued December 17, 1996.

Levine, E. Reduced Search for Macroblock Motion Partitioning in MPEG-4 Encoders

U. S. citizen.