Designed and led the implementation of an ads automation solution that automatically added AI-generated and in-house creatives to advertising campaigns based on supply strategy, reducing creative fatigue and improving CTR. The solution was integrated into TikTok Ads ad group creation flows and contributed to approximately 5% of TikTok creative revenue.

Designed the multi-DC server, storage, and traffic topology to support global serving and multi-site high availability.

Built similarity understanding infrastructure for video and image content to identify original, homogeneous, plagiarized, and remixed materials, contributing to a 7.6% lift in impressions, a 2.7% lift in clicks, and identification of 17% low-quality creators.

Designed and led the adoption of a Python-DSL and C++ runtime recommendation framework for recommendation and search scenarios, enabling 40–70% latency improvements for migrated workloads while significantly improving developer productivity.