Oracle Senior Applied Scientist

Oracle Health

Building LLM-based automation for EHR Message Center workflows — automated patient message drafting and clinician-facing summarization, from system design to production.

TikTok Senior Applied Scientist

Livestream Trust & Safety

Led a team building ML systems for livestream safety — age estimation and automated moderation, built around multimodal LLMs in multi-stage decision pipelines.

• Minor-safety recall pipeline (XGBoost → multimodal models → fine-tuned in-house MLLM): human confirmation rate 30% → 70%+, automated decisions 0% → 60%+ at 90% precision.
• Automated moderation framework: non-recommendation enforcement 30% → 90%+, interruption enforcement 0% → 35%, significantly reducing human moderation cost.

Yitu Technology Machine Learning Engineer

Applied AI

Led teams shipping applied AI projects across autonomous driving, retail analytics, and medical imaging.

• L3 autonomous driving perception: LiDAR detection with custom CUDA kernels (5×+ speedup); passed the Shanghai road-test licence exam.
• Customer behaviour recognition: multi-camera pose tracking at mAP 0.95 / 80 FPS; deployed in malls, banks, and airports.
• Children's bone age prediction: TW3 + deep learning, < 0.3-year deviation in clinical use.

Shanghai Jiao Tong University

M.S. & B.S. Computer Science

• M.S. (2015 – 2018) at the BCMI lab, advised by Prof. Liqing Zhang visual representation learning and interpretability of deep neural networks
• B.S. (2011 – 2015) Computer Science and Technology

Deep Part-Based Image Feature for Clothing Retrieval (ICONIP 2017) — L. Zhou, Z. Zhou, L. Zhang.