Role
I am currently working as a Software Engineer, ML Infra at HARP Research, Inc., since January 2025.
Project
- AdApt: A personalized advertising tool leveraging NLP, multimodal data, and automated pipelines for hyper-personalized messaging.
Responsibilities
- Lead developer of a style transfer system for personalized advertising, making messages feel more natural and improving engagement through semantic scoring and few-shot fine-tuning.
- Implemented text style transfer techniques for enhancing personalization and user engagement.
- Delivered an internal Streamlit tool to support business demos and unblock frontend/backend alignment, enabling faster stakeholder feedback.
- Refactored a shared text/image scraping framework, reducing response latency by $83%$ and Docker build time by $70%$ via modularization and targeted caching.
- Co-developing a Vue web app for the demo launch and using Streamlit for model validation and edge case testing.