Mr. Yifan Zhang

Software Engineering
Generative AI, LLM, ASR, TTS

Experience

Senior Software Engineer, QCRI 2013-present

Developed QATS speech recognition system, QAT Transcription and Translation system, live speech

recognition, and video search platform (NodeJS, React, Python, Django, C++, KALDI, MySQL)

Organize MGB Arabic ASR Challenge – provide technical support for MGB Arabic ASR Challenge; support

data preparation and test for challenge (Python, KALDI)

Created QCRI news translation system and led team to win BBC ‘Best in Show’ competition – create the

idea for QAT Transcribe and Translation system, led the development team for Hackathon (Bash, Python)

Implemented QCRI Arabic Transcription System with in-house ASR system, worked closely with Al Jazeera

Evaluate ASR and initiated in-house ASR system development (Python, Django, KALDI)

Consultant, QCRI 2012-2013

Consultancy on speech recognition for Al Jazeera – worked on EML speech recognition system, explored

KALDI for in-house speech recognition system (HTK, Bash, Python, KALDI)

Consultancy on architecture and design for BORAQ (later Jalees) EReader app (Objective-C)

Redesign and developed improved EReader app (Objective-C)

Backend Engineer, Wavii Inc, (a Google company) 2012-2013

As backend NLP Software Engineer, I work on the information retrieval component of the system, research

and implement new algorithms, improve system performance: Work on core information extraction

algorithms (Python, NLP, NLTK, R)

Work on components integrate backend and frontend information (Python, django, MySQL, Chef)

Improve overall system performance, implement internal requests (Python, AWS, MySQL, Chef, Pyramid,

Javascript)

Senior Research Engineer, SpinVox Ltd. (a Nuance company) 2008-2012

As a Senior Research Recognition Engineer in SpinVox, I have worked with the team to develop the

speech recognition from ground to be the core part of the company. I contributed in acoustic training, in

house development, experiments, and largely decoder development and optimization. Highlights are:

Analyze data and conduct experiments to improve core speech recognition system (Bash/Python/Perl)

Analyze and implement state-of-art techniques (vtln, gender, hlda, lexicon tree, combination) integrate

into current system (C++/C) Optimize both algorithm (Gaussian selection, computation cache, search)

and code (SIMD, cache, compilation) to achieve better performance (C++/C)

Design and implement features API, architecture such as shared memory, cached resource, multi-processsharing, exception-safe and etc. for our application (C++) Provide solutions, frameworks and supports for

system training, experiments, research and development in parallel environment (Bash/Python/Perl)

R&D Engineer, Virage Softsound (Autonomy, a HP company) 2005-2008

As Research and Development Engineer, I quickly integrated into a small team and developed Mandarin

speech recognition system, co-developed other languages such as Romanian, Arabic and Japanese

system, in house application development and software development, as well as research environment

management.

Highlights are: Build speech recognition system acoustic model and language model (Bash/Perl) Analyze

and evaluate data, models and techniques (Bash/Perl/Python) Develop in-house applications such as

text segmentation, dictionary generation, normalization and web crawlers (C/Python/Perl) Develop C

library for error handling, memory, data structure, threading and so on (C) Work on projects including IIVR,

transcript server (C++/Multi-threaded) Manage research environment of Linux and GridEngine (Bash)

TianChuang Software 2001-2002

As Lead Software Engineer, I led the development of bespoke Internet Information Management system

and mentor other graduate developers. Highlights are: design and develop Internet information

management system (ASP/SQL)

mentor teammates and integrate team member’s work into final product (ASP, SQL)