Alibaba DAMO Academy 2018-present
Chief Scientist and Senior Director of Language Technology Lab, DAMO Academy
"Make Language Local, Make Business Global"
Developed one of the first Chinese Large Language Models (LLM): PLUG and AliceMind, which evolve into Tongyi Qianwen, Alibaba's ChatGPT system, and achieved results surpassing human performances for the first time in tasks such as machine reading comprehension (MRC), visual question answering (VQA) and Chinese understanding (CLUE), which were highlighted in Alibaba's financial reports twice.
Applied LLMs in various Alibaba business scenarios and product, such as Dingtalk (Slack), Tmall Genie (Alexa), customer service chatbots, search, advertisement, translation, etc. and generated tens of millions of dollars in revenue
Built AliNLP, the NLP service platform that supports dozens of BUs within the Alibaba Group with trillions of calls per day, providing information extraction, sentimental analysis, natural language understanding, machine translation, OCR, language generation, embedding, etc.
Developed AI products to empower judicial, e-commerce, health, customs and other industry partners, enable clients with advanced AI solutions.
Led the development of ModelScope open source AI model community, the largest and fastest growing AI model community in China. Within 6 months of launch, it already hosted 800+ models, 2million users and 20 million model downloads.
Zhejiang University 2020-present
Adjunct Full Professor in Computer Science
Meta (Facebook) 2014-2018
Senior Tech Leader/Staff Research Scientist in Machine Translation, Applied Machine Learning and Places Data, Local
"Connect the world in everyone's language"
Develop cutting-edge technologies in machine translation, confidence estimation, language detection, user language modeling and natural language understanding, Reduce language barriers for 20 Billion people, built the best social media translation systems among 45+ languages.
Lead the research effort on machine translation and machine learning for Places Data, setting vision and agenda for the team, help oversee the execution and make sure the delivery of strong results with high impact.
Mentor junior team members; provide technical guidance to the team, and collaborate/communicate with cross-functional teams within the company and external collaborators so that the technologies developed at the translation team can have impact to other products and organizations.
IBM T. J. Watson Research Center 2006-2014
Senior Research Staff Member in Statistical Machine Translation
"Make Watson multilingual"
Develop cutting-edge technologies in statistical natural language processing and data mining to significantly improve IBM SMT system performances.
Drastically improved the accuracy of the IBM English-Chinese SMT system by 46% via better modeling of phrase translations.
Technical lead for the MT between Arabic and English in Meedan project, where MT technologies are used to help the communication between Arabic-speaking and Englishspeaking communities.
Lead contributor in the IBM RTTS real time machine translation Platform to offer high quality, multilingual (English<>Spanish, English<>Chinese etc.) chat and web translations services.
Major contributor in the IBM-led Rosetta team (including CMU, Stanford, JHU and UMD etc.) in the government-funded GALE project, responsible for improving word alignment accuracy, MT system performance and system combination.