Jin Han

• Collaborated with the Indian team to compute monthly Asia ROI, analyzing brands index to inform next month's marketing strategy.
• Implemented predictive algorithms to forecast beer ratings, identifying critical brewing process elements affecting taste.

Data manipulating

• Utilized the Selenium framework to simulate user actions for login, clicks, and pagination operations. Wrote automated web crawling scripts, including anti-scraping measures, successfully collecting over 20,000 skin images
• Conducted research on various OCR technologies to extract and batch process text from prescription images and medical reports, Results were used by the algorithm team to upgrade and refine the skin classification model, achieving a significant improvement.
• Designed and implemented data synchronization between two major internal databases using sqlalchemy, PostgreSQL plugin dblink, and cron jobs, enhancing data sharing efficiency across teams.
• Formulated and improved data cleaning, anonymization, and storage guidelines. Led the development of an automated data anonymization and storage platform, improving data management and processing efficiency.
• Managed the storage, backup, and retrieval of multimodal massive datasets for various product lines. Proficient in using NAS, minIO, and Alibaba Cloud storage services.

Data Mining

• Developed an evaluation metric system for the company's multi-disease prediction models.
• Independently completed a facial segmentation project based on Naive Bayes, with the model results in production use.
• Conducted AB testing analysis for pricing strategies on mini-programs and performed funnel analysis.
• Designed tracking plans for the annotation platform and wrote quality control analysis scripts using sqlalchemy and plotly.

Multimodal Annotation Platform Development

• Utilized computer vision, machine learning, and unsupervised hierarchical clustering algorithms to deploy lesion annotation graphical fusion, reducing annotation and case cost by 30%.
• Designed and optimized automated ETL processes for high-frequency annotation projects, leading the development of a one-click duplication feature for annotation projects.
• Led and managed the application of quality analysis modules on the annotation platform. Implemented online quality control for annotators, resulting in cost savings of 5%-20% per project for the company.
• Authored general-purpose script modules for data processing in various annotation platform workflows, shared across the entire data team, reducing response time and improving data processing efficiency.

• Cleaned 1 billion rows of raw data for CNBC network with Pyspark
• Built Predictive Regression Models on Pyspark with ml and MLlib packages
• Contributed to R&D tasks to enhance currently implemented forecasting methodologies
• Contributed to building data-processing algorithms and visualizations for dashboards and BI tools