● Extracted bond ratings from 1 million unstructured PDF data with 87% accuracy using an ensemble of LLMs

● Conducted comprehensive data analysis on extracted interest rates and issuance costs, providing actionable financial insights

● Developed and led a course covering machine learning topics and tools; offered to a class of 20 data scientists from the CIA

● Created a wide range of assignments, including JAX, NLP, LLM fine-tuning, and CV with PyTorch and Tensorflow

● Designed and constructed data pipeline to facilitate 1 million+ daily sales data extraction, transformation, and loading (ETL)

● Updated company data warehouse daily on Google BigQuery with VMs hosted on GCP

● Conducted time-series forecasting using ARIMA and XGBoost to predict demand for stock-keeping units (SKUs)

● Leveraged market basket analysis to uncover cross-selling opportunities from customer purchase patterns

● Automated personalized messaging to 1.9 million Shopee customers by developing a Python application

● Achieved intelligible speech synthesis using as little as 30 seconds of audio data when adapting to an unseen language; co-authored Few-Shot Cross-Lingual TTS Using Transferable Phoneme Embedding and accepted at INTERSPEECH 2022