End-to-End Data Pipeline and Sentiment Analysis for Market Insights – Gogoro Inc.

This is a project in collaboration between NTUDAC (National Taiwan University Data Analytics Club) and Gogoro (a Taiwanese electric scooter company). Its primary objective is to assist Gogoro's management and marketing teams in gaining insights into the current market landscape by analyzing consumer feedback and attitudes towards Gogoro on the internet. The findings will aid in making informed decisions for the next steps.

Due to the confidentiality of the project, the Github repo is kept private.

My Contribution

Developed an automated ETL pipeline for social media sentiment analysis and top keyword using GPT-3.5 API and HuggingFace NLP model, categorizing data by product features and month, saving 20+ hours of manual labeling weekly.
Built a dashboard with Python Streamlit and Plotly, integrating the pipeline to deliver insights with one click in under 60 seconds, while leading and training the team on API integration and front-end development.

Dashboard Demonstration Video

Final Presentation Slides

Git

Working with a team of seven, our team required a version control system and pre-checks. As a result, I gained extensive experience with Git.

ML & Pipeline

By building pipelines with the ChatGPT API, I gained valuable insights into the data science project lifecycle, including machine learning deployment and automated pipeline construction.

Project Management

The complexity of this project, along with the diverse skill sets of team members and time constraints, provided me with lessons in effective project management.