Publication

General Introduction

Wenlong is currently working as Principal SW Engineering Manager at Microsoft, leading the media AI team to develop and optimize the video effects pipeline for Microsoft Teams and Skype, including adjust brightness, soft focus, etc. 13+ years' experience in AI and SW covering image processing, NLP, machine learning, model acceleration,and etc., invented 49+ patents (including 19+ PCT), authored 10+ papers and led/participated in 35+ AI & SW projects. Initiator and creator of Intel's internal machine translation system, previously CV leader of Intel, Ecarx, Varian China and Digital Health China. Hosting PMP and LESS certificates.

Recent Highlights

- Released 'soft focus' and 'adjust brightnesss' visual effect features for Microsoft Teams in May 2022:
    What’s New in Microsoft Teams | May 2022
    How Microsoft Teams uses AI and machine learning to improve calls and meetings

Roles and Membership

- One of the Innovation Drivers for SOX department (~300 EEs) at Microsoft
- Regular MICCAI Member
- Previously innovation champion for around 60 EEs of Intel Flex GEAR region (China, Israel, India, and Ireland)
- Previously member of 3 Intel IP Committees: ‘Automotive, Drones, and Robotics’, ‘System Software and AI’ and ‘Camera, Audio, Video & Image’
- One of the main code implementors of China DRM protocol and CDTA

Working Experience

2021.11 - Now      Microsoft Corporation
                             Principal SW Engineering Manager
2021.7 - 2021.10  Digital Health China
                             Senior Image Algorithm Director
2019.6 - 2021.6    Varian Medical Systems
                             Senior Data Science Manager
2018.5 - 2019.5    ECARX Technology Co., Ltd
                             ADAS Algorithm Director
2008.1 - 2018.4    Intel Corporation
                             Senior Deep Learning Engineer
                             Senior Software Engineer
                             IT Flex GEAR Innovation Lead (~60 EEs)
                             Project Manager (2~10 people)
                             Member of 3 Intel IP Committees

Education Background

2022.9 - Now
- University:Tsinghua University
- Degree: PhD Candidate
- Major: Computer Science

2006.9-2009.3
- University:University of Shanghai for Science and Technology
- Degree: Master
- Major: Measuring Technology and Automatic Equipment

2002.9-2006.7
- University:University of Shanghai for Science and Technology
- Degree: Bachelor
- Major: Electronic Information Science and Technology

Wenlong's top 5 strengths

Futuristic, Deliberative, Restorative, Intellection, Maximizer

METHOD, APPARATUS AND SYSTEM OF RECOVERING AN OPERATING SYSTEM

A device, method and system for recovering an operating system may comprise a near field communication module; In some embodiments, the recovery manager may initiate a session between portable devices through a near field communication connection established by the near field communication module.

DISPLAYING AREA ADJUSTMENT

In an embodiment of the invention, a portable electronic device may conserve power while in a normal power mode. The display screen of the portable electronic device may be switched from an original screen mode to an adjusted screen mode. In adjusted screen mode, the display screen may display an adjustable displaying area that is reduced from an original size and it may include an inactive area sized to reduce display consumption of power.

CATEGORY-BASED KEYBOARD

Apparatuses, systems, media, and/or methods may involve a keyboard and/or a selection of an element thereof. The architecture of the keyboard configuration may be based on one or more categories, such as a hand category and/or a finger category. In addition, a touch area may be recognized, which may be formed between a user digit and the key of the keyboard configuration.
View ALL 47+ Patents

Deep Learning based ADAS PoC on Intel GO Platform

The 1st Object (car, person, truck, traffic lights and speed limitation signs) detection and road segmentation PoC/reference design on Intel Go (Car Branch: Deverton + A10) platform, for which we performed deep learning model training, customization and optimization, with Intel A10 FPGA acceleration. [2016.9-2018.4]

AI based Auto Contouring for OAR(Organs At Risk)s

Led team developed a 3D U-Net based AI model to auto contour OAR(Organs At Risk)s. It supports many tumor types and can reduce the manual contouring time of physician sharply. This can be deployed to hospital as an independent local service or integrate with multiple exisitng radiotherapy software [2019.6-2021.6]

Open source machine translation plus post-editing

Developed a machine tranlsaion engine based on Moses, with its corresponding web service API, windows client, and web page access. With these capabilities, it can be used to help to reduce cost for Intel's localization process, by using the model called 'MT + Post-editing'. This MT engine is being maintained and used for Intel internal for a long period until now. [2009.1-2010.12]
View ALL 35+ AI & SW Projects

Thought Vector

A thought vector is like a word vector, which is typically a vector of 300-500 numbers that represent a word. A word vector represents a word’s meaning as it relates to other words (its context) with a single column of numbers.

Machine Learning Algorithms used in NLP

Natural Language Processing is a very useful domain of using machine learning technologies, and has been developed for many years. Now devices are getting smaller and smaller, NLP becomes a possible better input/output way for small devices.

AI-complete

In the field of artificial intelligence, the most difficult problems are informally known as AI-complete or AI-hard, implying that the difficulty of these computational problems is equivalent to that of solving the central artificial intelligence problem—making computers as intelligent as people, or strong AI.
View ALL 233+ AI Notes
Four-dimensional computed tomography (4D-CT) has been used in radiation therapy, which allows for tumor and organ motion tracking through the breathing cycle. In this paper,we apply a deep learning based deformable registration network (RegNet) to estimate the deformation field for the given scan pair of 4D-CT. The proposed network was trained in an unsupervised manner without the need of any expert annotation.
View ALL 10+ Papers

National Fundings

吴伟棠,李咏梅,沈琳,杨文龙,李济时. 基于国际共识和人工智能实现鼻咽癌自动勾画和自动最优剂量学计划,香港大学深圳医院 (合作单位:瓦里安医疗),2021年深圳市基础研究面上项目,2021.8.16

Awards

Certificates

Recent Activities

- Attended MICCAI 2020 from Oct. 4th, 2020 to Oct. 8th, 2020 online
- Attended MICCAI 2019 in Oct. 2019 at Shenzhen, China
- Attended CES in Jan. 2019 at Las Vegas, USA
- Attended Intel PES IT Technical Conference in 2017 at Portland, USA
- Attended TAUS Conference in 2009 at Portland, USA

Open Source Projects

Winnie is a home robot project developed by me in my spare time started from July of 2017. She is composed by 4 wheels, 2 robot arms and a neck to move cameras with the head freely. The cost for its hardware is less than 2000 RMB. The initial intention of making this robot is to try different algorithms and collecting data more conveniently.

- Watching movies
- Listening to music
- Painting
- Traveling

Contacts

Wenlong's LinkedIn Profile

- Email: 290125097@qq.com

Feel free to get in touch or leave me a message!