Collaborative and skilled software/algorithm engineer with six years of experience in approaching software development challenges with a positive and solutions-oriented mindset.
I have expertise in AI software stack and possess significant experience in software infrastructure development, particularly in building distributed inference systems. Throughout my career, I have successfully completed multiple projects encompassing various areas such as object detection, object tracking, face recognition, social distancing estimation, license plate recognition, image stitching, keyword spotting, automatic speech recognition, and more.
At my company, I am known as the go-to problem solver who is always looking to innovate and find efficient solutions. I thrive on expanding my knowledge of coding and the challenge of being an algorithm engineer truly brings me passion in my work.
Senior Software/Algorithm Engineer
Taipei, Taiwan
[email protected]
C/C++
Python
Shell Script
CMake
Git
Docker
Doxygen
Swig (API for python from C++)
OpenCV
OpenCL
Linux
multi-threads
distributed computing
serialization/deserialization
CICD
data version control (DVC)
MLOps
PCA
Interactive Segmentation
Connected Component
Image Stitching
Direct Linear Transformation
Kalman Filter (Tracking)
SIFT
Hough
HoG
Image Deblurring
Depth Estimation
ISP
Framework Transformation
Data Augmentation
Transfer Learning
Model Pruning
Model Quantization
Performance Evaluation
Parameter Fine-tuning
Model : SVM, LeNet, AlexNet, VGG, GoogLeNet, SSD, YOLO, MobileNet, ShuffleNet, FaceNet, Xception, MatrixNet, CenterNet, CSPNet, M2Det, EfficientNet/Det
HPC
Keyword Spotting
Automatic Speech Recognition
Distributed Inference System
Social Distancing Estimation (Lidar)
Object Tracking
Object Detection
Automatic License Plate Recognition Optical Character Recognition (OCR)
Multi-Camera Object Tracking (Image Stitching)
Voice Keywords Detection
Face Recognition
Blur Detection
People Counting
Over the Wall Detection
Automated Optical Inspection(AOI)
Interactive Segmentation(for ROI)
Web crawler
1. Enhance the Image Signal Processor (ISP) algorithm, especially in 2D/3D noise reduction. Leverage cutting-edge techniques from both deep learning and image processing for a comprehensive improvement.
2. Tuning of parameters to achieve optimal image quality.
1. Established standardized software development process for SW AI team.
2. Refined the code of projects involving TinyML and TVM to enhance reusability and readability.
3. Designed and developed an AI Audio System from scratch, including the conversion of the program into C code for FPGA implementation, and integrated Keyword Spotting (KWS) and Automatic Speech Recognition (ASR) functionalities.
1. Develop a complex distributed inference system software product, starting from the conceptual design stage and guiding the project through all phases, from development to successful release of our first-generation product(NPU). (core libraries, python API, c++ API, and documents) [python, c/c++, , shell script, cmake]
2. Implemented CI processes for several modules. [python, shell script, git]
3. Developed multiple operational test flows for various OP. [c/c++, python]
4. Develop some passes for reconstructing the combination of OPs to make new model compatible with our chip. [python]
Reason for leaving: The company went bankrupt and I was forced to leave my position.
Develop a social distancing estimation project from the ground up, utilizing C++ programming language and incorporating both camera and lidar technologies. The project will be optimized for embedded devices. [c/c++, python]
Reason for leaving: Headhunter offered me a job with more competitive salary.
Led AI teams in developing computer vision algorithms for embedded systems, while also establishing standard operating procedures for training machine learning models. Notable projects include:
1. License plate recognition [c/c++, python]
2. Vehicle tracking [python]
3. Trash throw detection [python]
4. Customized augmentation tools [python]
5. Porting algorithms to embedded devices - Android Phone and JetsonNano.[c/c++]
6. Transferring models between different frameworks such as Darknet, ONNX, TensorFlow, SNPE, and TensorRT. [python, c/c++]
Reason for leaving: I was offered a new opportunity by a former manager who has recruited me to join ViewSonic.
Develop the computer vision algorithm for the IoT products. Implement and adjust the latest deep learning model for solving the object detection problems, and evaluate those performances. The project I've been in charge of are listed below. [python]
1. Object Detection/Classification.
2. Object Tracking.
3. Interactive Image Segmentation with GUI for intelligent ROI system.
4. Data Augmentation Tools with GUI.
5. Blur Detection on the image of the human face.
6. People Counting with GUI.
7. Establish an efficient standard procedure on training neural network models from data processing to model evaluation.
8. Over the Wall detection with GUI.
9. Web Crawler
Reason for leaving: The office of company in Hsinchu was closed.
Develop state-of-the-art algorithms in the field of the multi-camera system. I've been handled with the projects below. [c/c++, python]
1. Image Stitching.
2. Object matching for multi-camera.
Reason for leaving: The company went bankrupt.
Develop the algorithm on the system of surveillance cameras by utilizing image processing and deep learning. I've been in charge of the projects as below: [c/c++, python]
1. People Counting
2. License Plate Recognition
3. Big Data Analysis
I have published four conference papers about computer vision. I develop the innovative algorithm in image processing, as below:
1. Increase the efficiency of deblurring the high-resolution image with 30-50%
2. Image recovering and depth estimation in Multi-camera.
3. Recover the motion-blurred image by utilizing the accelerometer.
1. “Image deblurring via block-based non-blind deconvolution algorithm.” Computer Vision, Graphics, and Image Processing, Yilan, Taiwan.
2. “Image deblurring by accelerometers of mobile phones.” Computer Vision, Graphics, and Image Processing, Yilan, Taiwan.
3. “Efficient image deblurring via blockwise non-blind deconvolution algorithm.” International Conference on Information, Communications and Signal Processing, Singapore, Dec. 2015.
4. “Morphology-based disparity estimation and rendering algorithm for light field images.” IEEE International Conference on Consumer Electronics-Taiwan, Nantou, Taiwan, May 2016.