Python Engineer

Ting-Liang Huang

Corvallis, OR ***********@*****.***

linkedin.com/in/ting-liang-huang-70689117b/ github.com/LeohuangLeo CAREER EXPERIENCE

AU Optronics Corporation, Taiwan July 2020 – Sept 2020 Computer Vision Engineer [Raspberry Pi, Jetson AGX Xavier, Python, Pytorch, Linux]

• Designed auto-collected images system using Raspberry Pi and camera devices with Python script and labeled the images in YOLO format, which contain the images and the corresponding text files that stored bounding box axis;

• Implemented object detection model (YOLOv4) for tracking panels in a box and inferred well-trained model on Jetson AGX Xavier;

• Achieved 99% test accuracy in the production line and finally connected between PLC and the edge device to achieve machine automation.

Oregon State University, USA March 2020 – June 2020 Graduate Teaching Assistant [C++, Linux]

• Designed writing/lab assignments and mid-term for computer architecture class and provided students with clearly defined course objectives.

EDUCATION

Oregon State University, USA Expected graduation: Dec 2020 M.Eng. Electrical and Computer Engineering (GPA:3.69); Machine/Deep Learning, Computer Vision, DBMS, GPU/CPU Architecture, Signal Processing. National Taiwan Ocean University, Taiwan Graduated: June 2018 B.S. Electrical Engineering;

Linear Algebra, Data Structure, Algorithm.

Coursera Certificate

IBM Data Science Professional Certificate, TensorFlow for Artificial Intelligence SKILLS

Language: Python, C++, Matlab, R, JavaScript, HTML, CSS; Machine & Deep Learning: SVM, Decision Tree, KNN, PCA, Clustering, Scikit-Learn, YOLO, Tensorflow, Pytorch, OpenCV; Database: MySQL, Db2;

Analysis & Visualization: Tableau, A/B test, Matplotlib, Pandas, NumPy; Others: Embedded Systems, Git, Linux, Flask, Microsoft Excel. SELECTED PROJECT

Road Following AI Robot [Python, Pytorch, Jetson Nano, Motor Kit] June 2020 – July 2020

• The project aims to create a smart robot following a small racetrack using Nvidia Jetson Nano operating on Jetpack 4.3;

• Utilized gamepad controller and raspberry pi camera to collect input images and target datasets: speed and direction;

• Trained the collected datasets using the ResNet-18 model and transfer learning technique;

• Inferred a well-trained model on Jetson Nano to control Jetbot moving smoothly on track following the road. Recommendation for Company's Expansion Using DBMS [Python, MySQL, JS, HTML, CSS] Jan 2020 – March 2020

• The project integrates twenty data sources that size ranging from 50MB to 1.5GB and containing all kinds of information relating to different countries to implement ETL pipelines;

• Utilized Python to correct spelling inconsistency of the name of countries among different files and extract useful information using correlation function, P-value, and domain knowledge;

• Imported pre-processed datasets to MySQL server and designed ER Diagram for our database management system;

• Built a website to present ranking results from SQL queries and give recommendations for the company’s expansion. German Traffic Sign Images Classification [Python, Pytorch, OpenCV] Jan 2020 – March 2020

• The project is to classify 40 classes traffic sign images size vary between 15x15 to 250x250 pixels using DNN models.

• Implemented image augmentation techniques such as rotation, flipping, and blurring on images to enhance the robustness and the number of datasets;

• Resized the datasets and utilized an adaptive learning rate to largely increase time efficiency in model training;

• Implemented EfficientNet-B7 model and Ranger optimizer to obtain 99.87% in testing accuracy. Online Store Monetization Experiment Design [Python, A/B test] Sept 2019 – Dec 2019

• Built an interactive and scalable Python dashboard to measure the A/B test in-store purchase flow impact by identifying the confidence interval and calculating the statistical significance;

• Identified a strong correlation between user conversion and form of payment;

• Recommended the addition of a credit card option as an alternative method of payment (AMOP) to drive sales;

• Increased the entire online store revenue potentially by 10.7% according to A/B testing experiment result while the goal is 5%.

Decision Tree Ensemble for Mushroom Classification [Python, Scikit-Learn] Sept 2019 – Oct 2019

• Analyzed the datasets that contain the class (poisonous or edible) with 22 categorical features related to mushroom's properties and applied One-Hot-Encoding and Label-Encoder methods to convert the data into numerical values with totally 117 features and one target class;

• Implemented Bagging and Random Forest algorithms from scratch using Python;

• Evaluated training results with confusion matrix and achieved 0.98 recall score of class edible.

Contact this candidate