My Works
GenAI-Powered Data Enrichment Project: Saving 108,000 Hours and $5.4 Million
AI Generates 1.M Tables Descriptions
THE HOW
Feature engineering - using domain knowledge to create new input features from large dataset of metadata and contextual information
Hyperparameter tuning – adjusting learning rate and batch size to optimize performance
Used pre-trained generative model fine-tuning to generate descriptions based on retrieved data
MY CONTRUBUTION
Discovery: reviewed processes and systems, conducted stakeholder interviews, built personas to seize the opportunity, create user cases and detailed user stories.
Design: collaborated with data scientists and engineers to document functional requirements and specifications. Partnered with UX to develop prototypes and conduct usability testing. Defined roadmap and MVP.
Delivery: defined segment target launch strategy, developed user training documentation, post-launch monitoring, and communication strategy for all stakeholders.
ACHIEVEMENT STATEMNT
Successfully leveraged GenAI to enrich descriptions for 1.2 million data tables, saving an estimated 108,000 hours of manual effort and cost about $5.4 million.
ABOUT ORG
With 1.2 million data tables and 40,000 tables containing over 100 columns of unknown contents, inefficiencies in data management, duplication prevention, and quality control have significantly impacted productivity and increased costs.
DATE
2024
Project Gallery
