02 — work

Selected work

Live products I've designed and shipped, plus the data science and NLP work behind them.

research & data

01

Multi-Label BERT Classifier

PhayaThaiBERT + DNN for Jira issue tickets. Reduced labeling time by 80%. F1: 0.769. Published at IEEE ICCI 2024.

NLPPythonDeep Learning
02

Document Clustering Engine

Unsupervised clustering pipeline using BERT embeddings to group related documents for topic categorization at scale.

BERT EmbeddingsCosine SimilarityNLP
03

ETL & Analytics Dashboards

Scheduled ETL pipelines from multiple data sources. Interactive dashboards in Tableau, Apache Superset, and Streamlit.

PythonTableauStreamlit