新竹港式料理.jpg

Rocking Lai

Introduction 

I am a deep learning engineer as well as professional software engineer. I take part in various deep learning & computer vision projects, from software application to AI chip. Also have the experience of developing video surveillance system (full-stack). In addition, I strive for learning and constantly looking for ways to share my knowledge. I am the mentor of Coursera, which helps students in learning computer vision and machine learning.

Skills 

Scientific 

  • Deep Learning
  • Machine Learning
  • Computer Vision
  • Image Processing

Engineering 

  • C++
  • Docker
  • QT
  • Tensorflow
  • Caffe
  • Object Oriented Programming
  • Scrum

Experience 

Machine Learning Software Engineer, AMD, Sep 2021 - Present 

  • Optimize machine learning library based on AMD's GPU
    a. GEMM (general matrix multiplication)

Deep Learning Engineer, CVITEK, Oct 2019 - Sep 2021 

  • Spin off from Bitmain.
  • Mainly focus on deep learning accelerator (so called NPU or TPU) and stereo vision accelerator.
  • Deep learning accelerator
    a. Research and prototype the int8, int4 and bf16 quantization algorithm.
    b. Post training based mix-precision algorithm.
    c. Model compression flow (heterogeneous quantization).
    d. Design quantization flow in deep learning compiler (graph optimization, calibration, fine tuning, mix-precision).
    e. Implement frontend of deep learning compiler (high level optimization and lowering to low level IR).
    f. Inference simulator via high level IR.
    g. Co-work with IC designer to design the cmodel of AI accelerator.
  • Stereo Vision accelerator
    a. Prototype the hardware-friendly stereo matching algorithm.
    b. Co-work with IC designer to design the cmodel.

Imlxnw96k4pw5ecmasom

Deep Learning Engineer, Bitmain, Aug 2018 - Oct 2019

  • Mainly focus on deep learning accelerator (so called NPU or TPU and so on)
  • Research of deep learning algorithm for edge AI accelerator.
    a. int8 and bf16 quantization algorithm.
    b. Post training based mix-precision algorithm.
    c. Quantization-aware training flow.
  • Deep learning compiler
    a. Quantization tool (calibration, fine tuning).
    b. Inference simulator for high level IR (cpu & gpu).
    c. Co-work with IC designer to design the cmodel of AI accelerator.

Imlxnw96k4pw5ecmasom

Algorithm Engineer, ULSee, Jun 2017 ~ Aug 2018

  • Vision algorithm for robot (object detection, gesture recognition, posture recognition).
  • Facial landmark tracking algorithm.
  • Improve face recognition flow.
  • Driver fatigue detection, phone talking detection for ADAS (Advance driver assistance system).
  • Plan & design the face recognition system for various projects. (IP camera integration, video management, lead the scrum flow).

Ovuysyw4opw53isngtk5

Software Engineer, NUUO, Sep 2014 ~ Oct 2017

  • Design and maintained the Network video recorder (NVR). It is an embedded Linux, which can received video stream from IP camera, various type of recording, video analytic, third party integration.
  • Develop new features for NVR client, which can play live video, playback video, smart search video, event management...etc.
  • Develop SDK of NVR, provide a way for third party to integrate our NVR.

Xpq7pojdp65cn0wbuqnj

Volunteer 

Course Mentor, Jan 2017 - Feb 2019

  • Machine Learning, offered by Stanford University.

  • Fundamentals of Digital Image and Video Processing, offered by Northwestern University.

Education 

Master's degree, Computer Science and Information Engineering

National Chiao Tung University (2012 - 2014)

Bachelor's degree, Computer Science

National Chiao Tung University (2008 - 2012)

Publication 


  • Toward Community Sensing of Road Anomalies Using Monocular Vision, IEEE Sensors Journal (Volume:16 , Issue: 8 ) 2016
  • Vision-Based Road Bump Detection Using a Front-Mounted Car Camcorder, IEEE International Conference on Pattern Recognition (ICPR 2014)
Powered by CakeResumePowered by CakeResume