Jingyi Wang 王景祎

E-mail: Jingyi.Wang1903@student.xjtlu.edu.cn
Phone / Wechat: 18537820068
Address: IR724, South Campus, Xi’an Jiaotong-Liverpool University, Suzhou Industrial Park, Suzhou, Jiangsu, China

Education

Xi’an Jiaotong-Liverpool University | Suzhou, China
BEng Digital Media Technology | 2019.09 – 2023.06

  • 2020/21 University Academic Excellent Award (GPA Ranked 1st out of 55)
  • 2021/22 University Academic Excellent Award (GPA Ranked 1st out of 42)
  • Courses: Analog Electronics | Digital Electronics | Data Structure | Algorithm | Machine Learning
    Engineering Mathematics | Signals and Systems | Computer Graphics | Networking | Microprocessor | Communication | Human-Machine Interaction | Multimedia | Mobile Computing ……

Research Experiences

Supervised by Dr. Chun Zhao | Synaptic Devices Group | Xi’an Jiaotong-Liverpool University
Brain-Inspired Computation and Synaptic Transistor | Research Assistant | 2021.05 – now

  • Artificial intelligence is limited by the speed and energy consumption of the von Neumann computer. Synaptic transistor may be the cornerstone for building new type of brain-inspired computer.
  • Our research group studies the materials and processing of synaptic transistors.
  • I am focusing on the algorithm, application and simulation of brain-inspired computing based on our synaptic transistor device.
  • Modeled the transistors by measuring electronic characteristics (e.g., channel currents with gate pulses); determined the non-ideal factors including non-linearity, limited number of conductance levels.
  • Developed a plug-in algorithm for Pytorch to introduce non-ideal factors of our transistor into the weight update process when training a neural network.
  • Evaluated the performance of the transistor running neural network applications through training models such as chest CT image classifier based on ResNet18 considering non-ideal factors.
  • Developing a peripheral circuit platform with ADC/DAC for 3-terminal synaptic transistor array controlled by microcontroller, where neural networks may physically run on it.

XJTLU GMaster Robot Team | Deputy Captain | 2019.09 – 2021.08

  • Wined the second prize in RoboMaster 2021. RoboMaster, hosted by DJI, is a very famous robotic competition in China because of its large numbers of robots and team members.
  • Designed the electronic controlling systems for five robots. Completed following highlight features with STM32 microcontroller: self-balancing robot with Mecanum wheels (can move sideways); two-axis gimbal with shooting; auto-targeting enemies and movement forecast.
  • Developed vision algorithms for detecting the position of enemies in the camera. A mixed algorithm with traditional CV and Deep Learning approach is proposed. Automatic targeting is implemented through a combination of embedded systems and vision algorithm running on miniPC.

Supervised by Dr. Di Zhao | Advanced Systems Lab | Institute of Computing Technology, China Academy of Science
Brain-Inspired Computing Chips and Applications Based on RISC-V | Visiting Student | 2022.06 – now

  • Designing open-source arithmetic units (neuron models, gradient surrogate functions…) and extend corresponding RISC-V instruction set for accelerating spiking neural networks (SNN) based on Nutshell open-source processor.
  • Developing low-power ECG recognition application based on our RISC-V chip with SNN accelerator.

Embedded System Design for Environmental and Energy Fields | Research Assistant | 2021.05 – 2021.12

  • Designed an underwater networking alarm device for illegal fishing (electrofishing).
  • Developed the embedded software of photovoltaic maximum power point meter, which measures the I-V and P-V characters of photovoltaic, displays on screen, and determine the voltage at the maximum power point by charging a capacitor using photovoltaic from 0 volt.
  • Developed the embedded software of life tester for supercapacitors, which measures the life of supercapacitor by charging and discharging repeatedly.
  • Developed the embedded software of photovoltaic sun-chasing platforms, which allows the PV to face the sun vertically by driving three motors.

Supervised by Dr. Shengchen Li | Detection and Classification of Acoustic Scenes and Events Group | XJTLU
Algorithm Optimization for Embedded Deep Learning for Sounds | 2021.09 – 2022.07

  • Focusing on software and hardware co-optimization for AI applications runs on low-power edge microprocessors where the memory and CPU frequency are limited.
  • Deployed a real-time Keyword-Spotting system based on CNN on a STM32 microcontroller.
  • Evaluated the performance of different pruning methods and quantization precisions on real hardware.

Publications

  • All-in-One Metal-Oxide Heterojunction Artificial Synapses for VisualSensory and Neuromorphic Computing Systems
    Qihan Liu, Li Yin, Chun Zhao✉, Ziang Wu, Jingyi Wang, Xiaoran Yu, Zixin Wang, Wenxi Wei, Yina Liu, Ivona Z. Mitrovic, Li Yang, Eng Gee Lim✉, Ce Zhou Zhao
    Nano Energy, Vol. 97, 15 Jun. 2022, 107171. Doi: 10.1016/j.nanoen.2022.107171.
  • Hybrid mixed-dimensional perovskite/metal-oxide heterojunction for all-in-one opto-electric artificial synapse and retinal-neuromorphic system
    Qihan Liu, Li Yin, Chun Zhao✉, Jingyi Wang, Ziang Wu, Hao Lei, Yina Liu, Bowen Tian, Zhiyuan Zhang, Zishen Zhao, Ruofu Liu, Changzeng Ding, Yunfei Han, Chang-Qi Ma, Pengfei Song, Ivona Z. Mitrovic, Eng Gee Lim✉, Zhen Wen✉
    Nano Energy, Vol. 102, Nov. 2022, 107686. Doi: 10.1016/j.nanoen.2022.107686. [Free access link]
  • Keyword Spotting System and Evaluation of Pruning and Quantization Methods on Low-Power Edge Microcontrollers
    Jingyi Wang, Shengchen Li
    Submitted to 17th National Conference on Man-Machine Speech Communication. [Arxiv] [GitHub]
  • Digital Twin based Maximum Power Point Estimation for Photovoltaic Systems
    Kangshi Wang, Jieming Ma*, Jingyi Wang, Bo Xu, Yifan Tao, Ka Lok Man*
    19th International SoC Design Conference (ISOCC 2022), pp. 189-190. [Paper]
  • Digital Twin based Capacity Prediction of Ultracapacitors
    Kangshi Wang, Jieming Ma*, Bo Xu, Jingyi Wang, Yifan Tao, Ka Lok Man*
    Submitted to IIoTBDSC 2022.

Working and Project Experiences

iMotion Automotive Technology (Suzhou) Co., Ltd. | Vision Engineer Internship | 2021.09 – 11

  • Developed a vehicle key-point detection deep learning model based on human pose detection algorithms for 3D vehicle pose estimation.
  • CNN is used as the backbone and then Transposed Convolutions are used to generate heatmaps. Each channel of the heatmap indicates a location of a vertex of a vehicle.

Suzhou Vsdeep Intelligent Technology Co., Ltd. | Embedded System Engineer | 2022.03 – now

  • Developed an edge IoT embedded device based on STM32 which has LTE, Ethernet and 220V Electricity Testing modules. The devices are used in construction sites.
  • The device monitors the ethernet devices in subnet and Internet by pinging them and measuring delays; It also monitors 220V electricity status and records error (e.g., no grounding).
  • The device reports the real-time status and detected error to the server using 4G Network with our customized protocol via TCP. Information from all devices is available on webpage.
  • The embedded software I developed is stable enough and works on hundreds of devices.

Shenzhen DJI Co., Ltd, RoboMaster | Teaching Assistant | 2019.01 and 2020.01

  • Worked as an algorithm teaching assistant for the RoboMaster Robotics Highschool Winter Camp hosted by DJI in 2020. This camp aims to help more teenagers to achieve their dreams of being an engineer, including me.
  • Guided the campers to get start on the Robot Operating System (ROS), learn path planning algorithms, and finally complete the competition in a simulated environment.

Pat Pat Duck Project in Smart To-C Science and Innovation Training Camp | 2021.02

This camp encourages university students to find pain points in life, survey user needs, and use design thinking to design smart technology go-to-market oriented products. Our group tried to combain the IoT technology with emotion and mental in our demo product: Pat Pat Duck. They are two duck-shaped dolls, used to provide warm companionship to long-distance hot couples. If one pat his/her duck doll, another doll in long distance will vibrate itself to convey the emotions from the boy/girl friend. I complete the hardware and program of the demo, while my members did the user survey and the lovely appearance design.

Selected Awards and Honors

  • The 12th LANQIAO Cup: National Software and Information Technology Professionals Competition | Algorithm Competition | C/C++ Group
    National First Prize | 2021.06
  • RoboMaster Robotics University Championship 2021 | International Group
    Runner-Up | 2021.08
  • 2021 Future City Plan-Hackathon: Sustainable Urban Development Solutions
    Second Prize | 2021.11
  • “Looking for Physics in Daily Life” Video Competition
    First Prize | Watch | 2020.05
  • National Olympiad in Informatics Provincial (NOIP) 2018
    First Prize | 2018.12

Skills

Software: C, C++, Python, Java, Pytorch, Linux, OpenCV, OpenGL, ARM Assembly
Hardware: Embedded systems (STM32, ESP32…), FPGA (Xilinx), PCB Design (KiCAD)
Extracurricular: Illustrator, Lightroom, Traditional Photography