Speech and Language Technologies

BabelCentral, Lexington, MA, 12/2008–2/2009

Speech Recognition Engineer/Scientist. Developed speech recognition technology for a novel on-line language learning software application. Designed and built a data collection and speech transcription database under MySQL, specified and wrote a protocol for efficiently marking pronunciation errors and managed transcribers and data flow. Quickly learned and wrote scripts to drive two commercial speech engines (command-line tools and APIs). Created tools in Perl and Python to run batch recognition, update database with results, and generate statistics to analyze and optimize performance.

Nuance, Burlington, MA, 2005–2008

Research Engineer. Support acoustic and language modeling research for rapid, automated transcription of voice mail (large vocabulary, continuous speech recognition) application. Acoustic model training, optimization and evaluation experiments on computational grid engine (SGE / Linux). MySQL database development, Python tools and CGI scripts to support automated progress tracking and source control (SVN) of experiment parameters.

Senior Data Engineer. Manage and develop data resources and processing tools used to develop automatic speech recognition (ASR) technology for transcription . Built small-scale web platform for R&D, writing Perl and Python CGI scripts for form driven queries and programmatic updates to MySQL backend. Design and code custom audio and text processing procedures integrated with computational grid engine (SGE / Linux). Installed and manage team Apache and MySQL servers, and DBA web interface. Perform data analyses using custom scripts and SQL queries.

MIT Lincoln Laboratory, Information Systems Technology Group, Lexington, MA, 2003–2005

Research Engineer. Speech and language technology R&D under contract with General Dynamics Advanced Information Systems (formerly Veridian). Performed studies of SVM methods for channel robustness in narrow-band speaker recognition and cross-channel effects on speaker verification. Designed and built a speech corpus meta-data query service. Developed tools for projects using text-to-speech systems, machine translation and confidence scoring of speech recognizers.

iPhrase Technologies (acquired by IBM), Cambridge, MA, 2000–2001

Senior Software Engineer. Natural language understanding (NLU) and information retrieval (IR) technology development. Software development in Python, Perl and C++; grammar development; analysis and tuning of core search engine. Design and development of internal APIs, modules and classes.
  Invented autonomous domain tuning of proprietary IR engine. Created document clustering procedure amenable to pre-sales analysis. Automated Q/A for structured data search capabilities across data attributes. Wrote filters for structured text streams to generate schema and populate tables using Perl and SQL scripts.

Philips Speech Processing (formerly PureSpeech), Cambridge, MA, 1994–2000

Sr. Speech Scientist. Advanced speech recognition technology development. Parameter training of recognizers for natural speech interfaces deployed in commercial PC consumer and corporate telephony markets. Software development of core technology in C and of in-house tool suites in Perl. Provide expert knowledge for lexicons, phonetic classes, phonotactic rules and grammars, and enhanced telephony applications.
  Invented bootstrap method for phonetic segmentation and classifier training of foreign recognizers given only English segmentation. Devised on-line word bias adaptation which autonomously improves deployed recognition accuracy. Integrated advanced hybrid speech recognizer into parent company legacy technology. Significantly improved recognition accuracy, signal pre-processing, classifier training and parameter optimization.

Symbus Technology (acquired by TextWare, aka Captiva), Waltham, MA, 1993

Software Engineer. Developed speech processing software. Implemented wavelet analysis, wrote C++ class libraries for underlying pipeline architecture, and built GUI-based application using Windows API and MFC libraries.

Other Research and Development

Brigham and Women's Hospital, Dept. of Radiology, Boston, MA, 2001–2003

Medical web services and informatics

Lead System Designer. Design, build and develop web services for patient record and radiology information management, and client-server interfaces to resident third-party hospital information systems. Administer networked services including Solaris NFS, NIS, Apache HTTPD and modules, and development databases, source code control and issue tracking. Provide software support to UNIX administrators and PACS managers.
  Created radiology report filing server that submits HL7 message requests via TCP/IP or HTTP to hospital's RIS and replies appropriately. Created patient record web services with DICOM back-end interface to hospital's PACS broker. Built software development infrastructure, significantly enhanced intranet site and services, and developed methods/procedures for unit and regression testing.

Massachusetts Eye & Ear Infirmary, Cochlear Implant Research Lab., Boston, MA, 1984–1988

UNIX system engineering and management

Research Engineer. Developed system and scientific software for psychophysical investigations with profoundly deaf subjects, the first recipients of artificial hearing devices in New England. Designed and implemented the first high-speed, multi-channel, Digital Sound UNIX device driver. Built the laboratory, acting lab manager responsible for materials requisitioning, administration, mechanical and electronic design, fabrication and testing.

Honeywell Electro-Optics Operations, Lexington, MA, 1980–1982

Semiconductor physics

Assoc. Research Engineer. Solid-state device physics, electrical characterization of long-wave IR detectors at low temperatures, automated data acquisition and supervision of technicians.

Center for Adaptive Systems, Boston University, Boston, MA, 1989–1993

Research Assistant (1991–1993). Research of dynamical neural network models with feedback architectures. Designed and implemented simulation software in C (scripting in AWK and Matlab) for neurodynamical modeling of psychophysical data.

Systems Administrator (1989–1991). Managed complex scientific computing environment. Configured heterogeneous platforms (BSD and Sys. V UNIX, DOS/Windows and Mac) as networked file servers, and administered over 70 user accounts. Installed and maintained computer hardware, SunOS, X-Windows, software packages and tools.

Massachusetts Institute of Technology, Cambridge, MA

Applications Programmer, Project Athena, 1989
Systems Programmer, Brain & Cognitive Sciences, 1988
Instructor (C  programming), Information Services, 1987–1988

Boston University, Boston, MA

Ph.D., Cognitive and Neural Systems (1995). Dynamical neural network models for speech perception and motor control. Quantitatively accurate simulations of psychophysical data on perception of variable rate speech and temporal variations in natural speech production. Digital signal processing implementation of auditory peripheral models and neural models for formant feature extraction.

Course work in neural and computational models of memory, pattern recognition and learning (e.g. backprop and ART networks), adaptive motor control, speech perception and production, computer vision and visual perception; probability and statistics; digital speech processing; auditory physiology and acoustic phonetics.

University of Pennsylvania, Philadelphia, PA

BA, Biophysics (1980), cum laude, College of Arts and Sciences; neurobiology research and Middle Eastern archaeology.
BSEE (1980), Moore School of Electrical Engineering; EKN society, Dean's List; research on superconducting tunnel junctions.

Discrete Math, Data Structures -- Boston Univ. Metropolitan College, 1987
DICOM Basics -- OTech, January, 2002
Project Management Basics -- Partners career training, June, 2002
Design and Implementation of XML/XSLT -- Boston Univ. CEC, July, 2002

Publications

Awards and Patents