Project A: Human Labour Censorship Efficiency and Quality Control in Douyin(Chinese TikTok) – Utilised Python to implement unsupervised machine learning techniques for setting sensible workplace productivity goals, and boosting workplace effectiveness, which improved censors’ accuracy from 93 percent to 98 percent in 3 months. – Collected and wrangled 1GB of censors’ actions data stored in Hive tables to set a standard for employees’ productivity by K-means clustering, and analysed its rationality by visualisation. – Monitored censors’ quality data based on different business metrics by creating and maintaining dashboards, and produced daily and weekly quality reports to identify the latest problems.
Project B: Workforce Scheduling Analysis and Automation – Allocated censors into 19 different daily work schedules to best fit the number of hourly incoming cases for review, which minimised labour cost by 8 percent under the 15-min time lag constraint. – Constructed a linear programming model that predicts daily time lags on labour distributions under various business needs using Python, which allowed evaluation of the current scheduling. – Achieved automated monthly scheduling using linear programming by arranging censors and dates in a table with business requirements embedded into rows and columns constraints. – Presented results to management team and wrote reports detailing algorithm and implementation to present to 4 other bases