Zhouhao Zeng

As the founding member in Meta robotics team, designed and built robotics vision-language-action (VLA) models that achieved SOTA performance.

As the tech lead and founding member on Meta AI LLM Post-training team, 1. Led the Meta AI agentic modeling, enabling advanced multi-step multi-tool agentic capabilities (e.g. code execution, data analysis, multi-step search, custom tools) 2. Initiated and drove Meta AI reasoning workstream, advancing model's capabilities on real-world reasoning tasks 3. Built and productionized Meta AI content search (e.g. Reels) capability

As the tech lead on Instagram Ads delivery and ranking team, I initiated and led the pillar for growing and improving Ads personalization and performance for low signal users: 1. Low signal users cover >50% of Instagram users, including users who are impacted by digital ads data privacy regulation, and users who have no/low activeness in Facebook family of apps 2. Initiated & formulated the strategy, and tech led the execution on Instagram Ads’ investment on privacy-preserving ads system, ads measurement, and ads machine learning techniques 3. Led the end-to-end optimization for low signal users’ Ads experience and performance through ranking, delivery, targeting, demand, signal attribution, and new product experience

As the tech lead and founding member on Instagram Reels ranking team: - Built and applied advanced machine learning techniques for Reels ranking and personalization: 1. Developed and productionized Reels’ first user and content understanding features into ranking model 2. Collaborated with Facebook AI to apply counterfactual evaluation and reinforcement learning based algorithms for Reels multi-objective value model optimization 3. Initiated and achieved the high-complexity redesign and migration of Reels multi-objective value model formula with 10+ optimization objectives - Led multiple machine learning initiatives to solve the unique product challenges for video recommendation: 4. Built and productionized Reels’ first creation-based ranking model to predict the user creation intent and inspire user creation 5. Led the Reels growth in US with end-to-end recommendation system’s US optimization that achieved +50% stronger growth in US than global 6. Developed a scalable diversity and exploration framework to improve content diversity and user experience

As a research scientist on Instagram Explore ranking team, I led and built the recommendation for Instagram Explore Stories: 1. Supported the end-to-end tech stack including backend, retrieval, ranking, and content safety 2. Developed and applied advanced machine learning techniques to power the recommendation for Explore Stories, which directly drove +80% user growth in 6 months

As a machine learning research intern in recommendation algorithm group at Alibaba, I developed personalized recommendation algorithm for Xianyu (China’s largest second-hand trading platform): 1. Improved CTR prediction model through extensive experiments on feature engineering. Devised multiple historical related features that proved to be predictive and valuable on online AB test 2. Developed a ML algorithm for prohibited items detection based on LDA topic modeling of user comments 3. Explored listwise ranking algorithm that targets on whole recommendation list optimization

As a computational biologist and bioinformatician, worked on cutting edge research on immunology and genomics in collaboration with leading biologists from National Institute of Health (NIH) and University of Iowa: 1. Published three (two co-first authored) papers in Nature Immunology and three papers in Nature Communications 2. Developed and leveraged advanced machine learning and statistical approaches to decipher gene regulatory network underscoring immune function 3. Devised novel bioinformatics approach for multi-scale integrative analysis of functional genomic data (e.g. ChIP-Seq, RNA-Seq, Hi-C data) 4. Developed scientific software to analyze massive genomics data using innovative algorithms and data mining techniques