Hello, Danjie here!
I’m broadly interested in technical stuff. I got my PhD in theoretical physics, studying string theory and its application in cosmology and black hole physics. But lately I’ve been working on applying machine learning to improving product quality and user experience. In particular, I’ve worked on search and recommendation products in Google Search and YouTube.
I’m also interested in ML research and have been trying to spend more time on it. Some of the blog posts here are for tracking and sharing my own learnings in this field.
Experience & Education
Tech lead of Brand Connect, a YouTube platform for connecting content creators with brands. I built various LLM based products from ground up that are critical for YouTube’s success in the creator economy. I led the quality, modeling, and data analysis workstreams across Brand Connect.
Tech lead of shopping and inspiration related ranking components. Significantly improved various offline and online metrics. Proposed various initiatives in improving the inspiringness and diversity of Image Search.
Tech lead of API security. Designed and led the delivery of a brand new end-to-end API abuse protection service.
Designed and built new anomaly detection ML models for Google’s networking infrastructure that backs various products such as Google Fi.
Applied unsupervised ML models to financial and social media fraud detection. Responsible for the end-to-end process from model development, quality iteration and external client facing delivery.
ML for enterprise security. Designed and implemented a model for botnet detection.
Physics research in applying string theory in cosmology and black hole physics.
Publication
-
Hyperbolic Black Holes and Open String Production
Danjie Wenren (Stanford U., ITP and SLAC)
e-Print: 1709.03590 [hep-th] -
Heavy Fermion Production and Primordial N-Spectra
Danjie Wenren (Stanford U., ITP and SLAC)
e-Print: 1706.01612 [hep-th] -
Tilt and Tensor-to-Scalar Ratio in Multifield Monodromy Inflation
Danjie Wenren (Stanford U., ITP and SLAC)
e-Print: 1405.1411 [hep-th]