cv

General Information

Full Name Wenqiang Lai
Email wenqianglai@hotmail.com
Languages English, Italian, Chinese (Mandarin)

Education

  • Oct 2021 - Oct 2022
    MSc in Applied Machine Learning
    Imperial College London, London, United Kingdom
    • Graduated with high Merit.
    • Awarded Distinction grade for the Master thesis.
  • Sept 2018 - June 2021
    BEng in Mechatronic Engineering
    University of Manchester, Manchester, United Kingdom
    • Graduated with high First-Class Honours.
    • Awarded 83% for the Final Year Project.

Publications

  • IEEE EUROCON 2023
    W. Lai, Q. Yang, Y. Mao, E. Sun and J. Ye, "Knowledge Distilled Ensemble Model for sEMG-based Silent Speech Interface," IEEE EUROCON 2023 - 20th International Conference on Smart Technologies, Torino, Italy, 2023, pp. 117-122, doi: 10.1109/EUROCON56442.2023.10198974.

Work Experience

  • Jul 2021 - Sept 2021
    Software Engineer Intern
    Huawei Technologies Co., Ltd., Dongguan
    • Worked in the R&D team to deliver 5G signalling system using C++ based on Huawei cloud service engine.
    • Ran meetings with senior engineers and managers regularly to ensure efficient development.
    • Wrote scripts in Lua for policy management module (used less than 50% of allocated time).

University Projects

  • Jan 2022 - Sept 2022
    Fall Detection using a Networked UWB Radar System
    • Delivered a real−time indoor fall detection system based on deep learning methods.
    • Generated a dataset consisting of fall/non−fall samples from 10 subjects.
    • Evaluated popular machine/deep learning methods (e.g., ResNet) in terms of classification accuracy, inference time and memory footprint.
    • The best−performing method achieved a test accuracy of 98.3% with an inference time of 4.63 ms and model size of 141 KB.
  • Oct 2021 - May 2022
    Silent Speech Interface based on sEMG Sensors
    • Built a robust and affordable (less than £100) silent speech interface (SSI) from scratch.
    • Generated a dataset consisting of EMG data samples from 5 subjects.
    • Trained popular machine/deep learning methods and evaluated their performance.
    • Best−performing model achieved 86.1% test accuracy on both PC and microcontroller.
    • Ranked first among all groups from the same course.
  • June 2022 - Sept 2022
    Knowledge Distilled Ensemble Model for sEMG−based Silent Speech Interface
    • Extended previous work on SSI by using more sophisticated sEMG sensors to classify 26 NATO phonetic words.
    • Proposed a deep learning method, which distills the knowledge from an ensemble voting classifier consisting of multiple 1D ResNet18.
    • Best−performing method achieved 85.9% test accuracy.
    • Awarded best student paper in IEEE Student Paper Contest at 2022 IEEE ACDS conference.
  • Jan 2022 - Apr 2022
    Human−centred Robotics: CareBot
    • Developed a robot to provide daily assistance (sending fall alert and help finding objects) to the elderly.
    • Developed and deployed navigation, fall detection & object finding modules on Pepper.
    • Pre−trained YOLOv4−tiny and MoveNet used to perform object detection and fall detection due to their superior speed.
    • Used ROS to provide inter−module communication & distributed computing, MongoDB for data storage and Docker to ease development process.
    • Presented the work in a live demo and compiled a technical report; awarded the highest grade amongst competitors.
  • Jan 2022 - Apr 2022
    Self−Organising Multi−Agent Systems: Simulation of El hoyo
    • Investigated the social behaviour of intelligent agents under resource constraints.
    • Designed and implemented a type of agent based on reinforcement learning (Policy Hill Climbing) in Golang. (GitHub)
    • Resulting agentbeat all other types of implementation w.r.t. self−organising ability.
  • Sept 2020 - Apr 2021
    Deep Learning for Human Activity Recognition Optimised for Microcontroller
    • Delivered an end−to−end system capable of recognising 6 human activities using wearable inertial sensors.
    • Built & evaluated a set of CNN−based models using TensorFlow.
    • Compressed the best−performing model using quantization, making it 14.6x faster and 3.9x smaller.
    • Successfully deployed the model on an Arduino with only 1MB CPU flash.
  • Sept 2019 - Mar 2020
    Autonomous Driving Embbeded System
    • Delivered a line−following buggy from scratch.
    • Developed the control system based on PID algorithm to compensate sensor errors.
    • Regretfully, this project was interrupted by COVID19 pandemic.

Honors and Awards

  • 2023
    • Top 5 Student Paper, IEEE Region 8 Student Paper Contest [Link]
  • 2022
    • Best Student Paper, Student Paper Contest at Imperial College [Link]
  • 2018
    • Gold Award, UKMT Junior Mathematical Challenge