Philip LEE Hann Yung

Designed Python web scraping scripts to extract textual content from 5 websites using Selenium. Conceptualised ELT pipeline for indexing, archiving, field extraction, and data cleaning to build high-quality, structured datasets.

Interfaced with Twitter API to build large-scale geo-tagged datasets (4 mil tweets) with ETL pipeline. Developed NLP-based geolocation models with composite feature engineering to predict user locations based on tweet features.

Utilized JavaScript and Google Earth Engine to aggregate biodiversity data on specific regions from public datasets. Conducted geospatial data analysis and generated visualizations using Python, R, geopandas, and matplotlib.

Conceptualised robust computer vision prototype using Python, OpenCV and PyTorch for real-time end-to-end social distancing and mask detection on network camera and low-power 15-watt Jetson device.

Investigated and trained 3 light-weight AI submodules, optimising for computation speed while maintaining good accuracy. Built and containerized system as Docker image for system reusability and deployment.

Utilised and trained Generative Adversarial Networks (GANs) for image-to-image translation tasks to augment existing image datasets to different domain, reducing model bias in semantic segmentation and other tasks.

Adapted multiple services and technologies (NGINX, PostgreSQL, MQTT, and Python) as the project stack on Linux-based system. Setup services on workstation server and integrated various sub-modules for team members.

Designed and implemented database logic design; incorporated dashboard app components using ReactJS.

Refined and debugged software, including creating Python automated scripts to capture and process real-time stream data from Bluetooth Gateways and IP Cameras.