Responsibility: led a Research and Development team including 10+ back-end and ML algorithms software engineers.
Developed a cloud platform and data mining system for automotive manufacturers;
Data Analysis: driving behavior analysis, transportation demands prediction, vehicle intelligent dispatching, large-scale trajectory analysis.
Technical stacks: ML/DL algorithms, Spring Cloud, Spark, Hadoop, Kafka, Hive, HBase, ElasticSearch, MongoDB, Redis, MySQL
Deep Neural Network Based Voice Activity Detection
DNN-based VAD has demonstrated good performance and has shown strong potential in practice. We use this machine learning method to optimize the VAD module of our speech engine.
A speech semantic analysis engine running on the mobile side
In practice, by analyzing the user experience feedback from our users, we may discover some bad cases of semantic recognition. These cases contain domain error, outdated POI/Music data, and polyphonic character. For example, the query spoken by user comes from navigation domain is recognized as music domain. Although we can optimize or fix some bad cases by training data and then update the semantic model in the cloud service, it is time-consuming. To quickly fix these bad cases, we develop a semantic analysis engine running on the mobile side.
Speech Recognition Android Application
We designed about 100+ voice commands for our driver users. I developed the application framework and some features related to voice commands of navigation and music.
Developed a C++ cross-platform map engine(OpenGL ES 2.0) for Android/iOS/Unity3D OS platform which is mainly used in QQ Augmented Reality(AR) red packet online game, PokemonGo-like mobile game; implemented some core feature like 3D-building rendering and route navigation; optimized the GPU performance from 20~40FPS to 40~60FPS.
“Zero to One”: led my team to create an innovative product named Tencent Automotive Services(TAS) which is an Android automotive operating system/In-Vehicle Infotainment(IVI), running in the head unit in Internet-connected cars.
Established and managed an Android software engineering team contains about 10+ engineers.
Designed and implemented the system software framework, developed core features and optimized the performance.
Baidu Maps Android Application Software Team, Team Leader, 7-11 engineers (http://map.baidu.com)
Contributed to the full mobile application development lifecycle from July 2012 to April 2014(Ver. 3.0-Ver. 8.x). Helped the market share of Baidu Maps services increasing from nearly 30% to 71%(Baidu Maps App owned about 326 million monthly active users up to 2015). Designed and implemented the framework for Baidu Maps Navigation Application; developed some core features in an efficient, maintainable and reusable design pattern; Created the first Android GPS navigation SDK of China in 2013, and it owned 100+ application users up to 2015.
Baidu Maps Navigation Map Engine Team, Team Leader, 6 engineers
Familiar with the technical mechanisms of map engine and optimized the performance via rewriting file format, upgrading algorithms and fixing some logic errors for online service: 30% reduction in file size, 15% reduction in time-consuming to load map blocks, nearly 60Mb reduction in memory-consuming; Led the team to develop an offline toolchain to transform the MID/MIF format raw data to readable format data for Android/iOS mobile applications.
Baidu Taxi-Hailing App project, Team Leader, 6 engineers
Developed smartphone applications for taxi-hailing service including a passenger-side component integrated into Baidu Maps App and a driver-side app. The achievements of this project gave technical support for Uber to integrate its service into Baidu Maps app. (in 2014, Baidu invested Uber, and they collaborated to expand Uber's presence in China to compete with DiDi Chuxing)