I was the Group Manager for the Applied Research and Engineering teams for the Trust, Safety and Integrity Intelligence team in Facebook Marketplace.
Facebook Marketplace now has one billion users. I led an organization of multiple interdisciplinary teams of research scientists and software engineers that use state of the art AI to safeguard these billion users and help them build trust on Marketplace. These teams were responsible for detecting policy violation on marketplace, as well as model complex behavioral analysis and intent understanding to detect fraud and scam on marketplace. To create a safe and healthy ecosystem on marketplace, the teams use state of the art Machine Learning like multi-modal models, reinforcement learning, active learning, graph neural networks, clustering methods as well HCI techniques to effectively leverage high‐judgement decision‐making where we cannot apply automation.
I directly supported engineering managers, TLMs and senior ICs in this role. My team is composed of Research Scientists, ML engineers and full stack engineers.
Patent: Ranked Pruning of Data Set to Train Machine Learning Models [Link]
Publications:
Mind the pool: Convoluational neural networks can overfit input size (ICLR 2023) [Link]
A Tour of Visualization Techniques for Computer Vision Datasets, IJCAI 2022 Workshop on Explainable Artificial Intelligence (XAI) [Link]
Improving Scammer Detection on Marketplace (in submission)
Learned Temporal Aggregations for Fraud Classification on E-Commerce Platforms (WWW 2023) [Link]
Training with Low-Label-Quality Data: Rank Pruning and Multi-Review (DMLR - workshop at ICML 2023) [Link]
co-located with KDD '22) [Link]