Chenyu Tsai
A data engineer and cross-domain practitioner with a passion for programming, linguistics, and design thinking. My primary research areas include explainable NLP systems and natural language understanding. I am dedicated to the development of robust data systems.
Additional research interests encompass human-computer interaction, user experience, design thinking, and writing.
Data Engineer
Work Experience
Junyi Academy, Data Engineer, Oct 2022 to Present
- Streamlined DevOps processes to enhance efficiency, resulting in a 95 % reduction in development iteration time, thereby accelerating product delivery.
- Successfully implemented cost management strategies, optimizing resource allocation and reducing storage expenses by 80 %.
Zigzag Analog, Data Scientist, Aug 2020 ~ Sep 2022
- Assisted clients in building data-driven business infrastructures, utilizing existing data and Tableau to create dashboards providing product insights, marketing information, and business goals.
- Designed and implemented A/B experiments for clients' digital funnels, improving the conversion rate by 11% and reducing churn by 6%.
- Leveraged SQL, cloud service Linode, and LINE to construct a digital platform monitoring system for clients. As a result, the response time to customer complaints was reduced from an average of 3 days to 2 hours, and the customer complaint/compensation ratio decreased by 27%.
- Built a sales data model and generated insights for operations and manufacturing, leading to a reduction in manufacturing costs by 7 % and a revenue increase of 5 %.
- Assisted client operations, finance, and sales in establishing a data-driven culture and SOPs.
Garden Villa, Special Assistant to General Manager, Jul 2015 ~ Aug 2018
- Cooperated with the General Manager in recruiting, training, payroll processing, performance evaluations, promotions, and employee release processes.
- Coordinated with the General Manager in planning short and long-term projects, budgets, expense controls, schedules, and manpower.
- Implemented quality and productivity objectives to achieve company goals.
- Supervised staff and controlled merchandise.
- Ensured company policies and procedures were adhered to.
Education
National Chengchi University, Master of Science（MS）, Digital Content, 2018 ~ 2020 GPA: 4.2/4.3
- Served as the Head Teaching Assistant for Artificial Intelligence and Design Thinking, mentoring 50 students to develop and execute 8 AI projects successfully.
- Achieved 1st place among 70 teams in the TRANS ACTION competition, collaborating with 6 teammates to develop a service that significantly improved the environment for migrant workers in Taiwan.
National Chengchi University, Creative Design Program, 2013 ~ 2014
National Chengchi University, Bachelor of Arts（BA）, Chinese Literature, 2011 ~ 2015 GPA: 4.0
Publication
- Modeling Meta-Explainability of Natural Language Inference
- NCCU 2022 Doctoral and Master's Thesis Award
- Published "Evaluating Model Explainability with Human-Involved Test," providing the Explainable Artificial Intelligence community with cutting-edge insights.
- In the Natural Language Inference scenario, identified that attention mechanisms do tend to attend key information after training with different but related tasks.
- Retrieved explanations for the NLI task from attention weight, analyzed the meta-explainability of different models and methods..
- Compared the differences between how models and humans create explanations, evaluating human trust and preference for the model explanation.
- Github Link
Projects
Below are some of my project experiences. Most projects combine information technology and user research/experience to better meet users' needs.
Comment Monitor System
Achievements:
- Reduced response time from an average of 3 days to 2 hours without increasing labor costs.
- Decreased the customer complaints/compensation ratio by 27%.
Methodologies:
- Utilized selenium and platform APIs to periodically obtain reviews.
- Deployed applications via the cloud service platform Linode and Docker.
- Integrated the LINE Bot API to send notifications to customer service staff.
Financial Data Analyzing System
Personal Development Project
Achievements:
- Selected financial analysis methods and data reported.
Methodologies:
- Deployed distributed scraper via Celery and RabbitMQ.
- Built RESTful API with Fast API.
- Deployed application via cloud service platform Linode and Docker.
- Wrote automated tests, maintained GitLab CI/CD, and managed version control.
- Built SQL database.
- Analyzed financial document with NLP sentiment analysis.
Mygran
- Secured 1st place in TRANS ACTION AWARD 2018, a competition that merges cross-domain knowledge and user experience.
- Proposed a service to tackle issues arising from the growing needs of migrant workers.
- As a core member and researcher, I participated in the entire research process and contributed to the application design.
Skills & Knowledge
- Explainable Artificial Intelligence
- Experienced in Visualization, Influence Methods, and Example-Based Explanations.
- Natural Language Processing
- Competent in Natural Language Understanding (NLU), Question Answering, and Text Classification.
- ML and Data Engineering Frameworks
- Skilled PyTorch, Numpy, Pandas, Scikit-Learn, and Airflow.
- Data Related Skills
- Proficient in GCP services (Bigquery, CloudStorage), dbt, RDBMS, SQL, Docker, Git, Tableau, and Product Analysis.
