Building a chatbot api on Microsoft Teams for interactive management using Node.js.

Fine-tuned the state-of-the-art natural language processing (NLP) model BERT to over 80% accuracy in the task of screening resumes written in Japanese, a 10% increase upon the previous model.

Built the pipeline for training, tuning and deploying the model on the API endpoint that has served over 10,000 resumes with Docker, Sagemaker and Tensorflow Serving.

Built an interactive dashboard visualizing automated process mining data using R and Shiny.

Achieved 92% accuracy on predicting annual water usage tendency in Germany using LightGBM.