Research • Two and half years experience in NLP/NLU research and development, including annotation definition, data tagging, model training and accuracy improvement. • Research on word segmentation for Mandarin. • Seven years of research experience in speech recognition and adaptation, voice conversion and text-to-speech synthesis. • Solid knowledge of statistics, machine learning and pattern recognition. Software Development • Over 10 years of experience with whole software development life cycle. • Proficient in C/C++ programming for algorithms implementation. • Proficient in Perl/Shell scripts for text processing and automation. • Extensive experience with Unix, Linux and Windows operating systems. • Hands-on experience with database and SQL. • Working experience with version control tools such as Perforce, Git and SVN. • Excellent trouble shooting and bug-fixing skills. Awards • Grand prize winner, champion of Nuance Innovation Challenge, Montreal 2013. Project for the Competition: One Person’s Theatre (An Application of Voice Conversion with Prototype Demo) Languages • Native speaker of Mandarin; fluent in English and Japanese. Doctoral Thesis: “Adaptation of Acoustic Models for Speech Recognition with Focus on Intra-Speaker Variation”

Sr. Natural Language Understanding Engineer (Sr. NL Research Scientist since 2013/10) @ * Research & development on NLU models for various products * Post-ASR Error Correction, using Pinyin Fuzzy Matching to correct ASR recognition errors which were caused by homophones and neighboring phones. * Research on incorporating word segmentation into NLU framework in order to reduce NLU tagging errors, which were often seen in Chinese and Japanese languages where there is no word delimiter between characters. From February 2013 to Present (2 years 11 months) Senior Consultant @ Skills: Object-Oriented Perl, Shell, CPAN, CGI, Catalyst, Template Toolkit, HTML, DB2, Sybase, SQL, XML, JavaScript, jQuery, JSON, Ajax, GWT, Java, Servlet, Apache, Linux, Andrew File System. Projects: Web development, from frontend user interface to backend module programming. * Enterprise Computing Request Portal System, which automates validation and approvals for user's requests, and auto-execute high volume tasks. * Grid Host Monitoring Daemon, which monitors health status of 20,000 grid hosts in different data centers and conducts regular maintenance. * Grid Host Audit, which deals with hosts with problems by performing remote reboots, rebuilds, and power-cycles; generates tickets to operation team. * Product Lifecycle Management System, which serves as a centralized repository for tracking products throughout their lifecycles from Acceptance Testing through to End of Support. From May 2010 to February 2013 (2 years 10 months) Postdoctoral fellowship @ Project: Voice conversion system for Cybit Technologies Inc. Skills: Speech recognition, text-to-speech synthesis, Linux, Unix, C/C++, STL, Perl, Shell, OpenMP, digital signal processing, statistical theory and methodology, model optimization algorithms. * Designed and developed a transform-based voice conversion system on Unix/Linux, using C/C++, Perl, Shell script and speech processing tools. This system consists of modules: (1) Feature extraction (extracting Mel-Cepstrum coefficients from speech waves using FFT/DFT algorithms and Mel-scale frequency filter banks); (2) Dynamic time warping (matching two given time series of parameter vectors); (3) Feature clustering (using k-means and maximum likelihood estimation algorithm to cluster parameter vectors); (4) Transform estimation (estimating linear transform between two speakers using linear regression models); (5) Signal restoration (using IDFT to restore speech waveforms). *Built large vocabulary continuous speech recognition systems on speech corpora using C/C++, Perl, Shell and Hidden Markov Model Toolkit. *Conducted research on discriminative training for large vocabulary speech recognition using second order cone programming for optimization. *Proposed a model-mapping approach to voice conversion by integrating speech recognition with speech synthesis techniques. Developed a baseline voice conversion system and implemented the approach. The system consists of modules: (1) Feature extraction from speech wave using Hidden Markov Model Toolkit and Speech Signal Processing Toolkit; (2) Gaussian mixture model estimation using expectation-maximization (EM) algorithm; (3) Model searching using probability of multivariate Gaussian distribution and normalized Euclidian distance; (4) Speech synthesis using HMM-based speech synthesis techniques and mel-frequency log spectrum approximation (MLSA) filter From January 2007 to April 2009 (2 years 4 months) Software Engineer/Manager @ After graduated from University, I worked at a software company in Dalian, China. As a software engineer, I was sent to the parent company in Japan, to develop commercial software for Japanese companies (anybody remember COBOL? Yes, that's what I used at that time!). Two years later I returned to China, worked as a Sr. engineer, sometimes on hardware but most times on software. Later on I led a department focusing on developing systems in financial industry , for four years. After that I went to Japan as a Sr. software engineer again. One year later I enrolled in Graduate School of University of Tokyo. When I came out of the school, I had two degrees in hands, and five years memory left behind in the campus... From July 1989 to October 2005 (16 years 4 months)

Ph.D, Information and Communication Engineering (Computer Science) @ University of Tokyo From 2000 to 2003 Masters, Information and Communication Engineering @ University of Tokyo From 1998 to 2000 Bachelor's degree, Computer Science @ Dalian University of Technology From 1985 to 1989 Baojie Li is skilled in: Automatic Speech Recognition (ASR), Text-to-Seech (TTS), Natural Language Processing (NLP/NLU), Part-of-Speech Tagging (POS Tagging), Machine Learning, C, C++, Perl, Linux, Unix, Python, SQL, Java, Algorithms, XML