Yu Xiang

Yu Xiang

Applied Scientist

Amazon

Biography

Yu Xiang got his PhD degree in Physics at Rensselaer Polytechnic Institute in 2019. He is currently an Applied Scientist at Amazon working on the state of the art LLM applications in the Amazon Smart Vehicle team.

Download his resumé.

Interests
  • Large Language Model
  • Natural Language Processing
Education
  • Ph.D. in Physics, 2019

    Rensselaer Polytechnic Institute

  • B.S. in Physics, 2014

    Wuhan University

Experience

 
 
 
 
 
Amazon
Applied Scientist
May 2021 – Present Seattle, WA

Responsibilities include:

  • Created a one-stop solution to the long-term forecasting problem with highly sparse inputs by designing a novel multivariate encoder-decoder attention model with two-head outputs.
  • Setup the data pipeline and deployed the models on native AWS to retrain and make predictions on any given cadence, while continuously monitoring the input data using a model drift detector.
  • Designed new features and expanded the existing ones to provide personalized recommendations to the drivers via Alexa by utilizing knowledge graph built from human annotation.
  • Improved the recall metrics of DNN models by 8% using state-of-art natural language processing technologies to help reduce the customer friction.
 
 
 
 
 
CGG
Seismic Imager
Jan 2020 – Feb 2021 Houston, TX

Responsibilities include:

  • Provided optimal quality control for the input seismic time series data by querying and analyzing information from billions of seismic records with SQL and Hadoop/Spark big-data tools.
  • Processed petabytes of seismic image sets by separating noise from signal using sparse representation, correcting artifacts with convolution/correlation, and applying interpolation with compressed sensing techniques.
  • Created high-fidelity 3D seismic images and velocity models by solving the optimization problem to minimize the square loss between model and data through an iterative gradient descent method.

Talks

2D Materials in Real and Reciprocal Spaces
Complimentary AFM and RHEED studies for real and reciprocal space analysis.

Publications

Quickly discover relevant content by filtering publications.
Searching for Circular Photo Galvanic Effect in Oxyhalide Perovskite Bi4NbO8Cl
Searching for Circular Photo Galvanic Effect in Oxyhalide Perovskite Bi4NbO8Cl

Abstract The ability to reconfigure spin structure and spin-photon interactions by an external electric field is a prerequisite for seamless integration of opto-spintronics into modern electronics. In this work, the use of electric field on the tuning of circular photo galvanic effect in a quasi-2D oxyhalide perovskite Bi4NbO8Cl is reported.

Monolayer MoS2 on sapphire: an azimuthal reflection high-energy electron diffraction perspective
Monolayer MoS2 on sapphire: an azimuthal reflection high-energy electron diffraction perspective

Azimuthal RHEED study about monolayer MoS2.

Contact