About me

I am an Applied Scientist at Amazon Last Mile Map Science Team, specializing in Computer Vision and Multi-Modal Machine Learning.

Currently, I am developing computer vision multi-modal large language models (VLLMs) for remote-sensing applications to serve Amazon’s last-mile delivery service. My work involves leading map feature extraction projects for the NA and EU region, and optimizing/correcting Amazon's map database to reduce logistic costs.

Education Background

Ph.D. in Mechanical Engineeringat Carnegie Mellon University (July 2020 – Jan 2025)

M.S. Research in Mechanical Engineering at Carnegie Mellon University  (Aug. 2018 to May 2020)

B.S. in Mechanical Engineering from the University of California, San Diego (Sep. 2014 to Jun. 2018)

News Letters

Selected Research

Large Polygon Language Model for Efficient High Quality Feature Extraction

  • Patent-pending VLM with novel coordinate tokens and redesigned vocabulary encoding spatial correlation, customized SFT and reinforcement learning.

  • Innovative dual-purpose inference prompting, serving as “grammarly” for database.

  • K. Qian, Y. He, M. Moustafa. Vision-Language Models for Building Polygon Extraction from Satellite Imagery. Under review.

Transformer-based Feature Extraction with Next Token Blurring

Novel High-order Algorithm on 3D Spline for Modeling Biological Neuron Growth

ML Future Frame Prediction (Video) using MetaFormer Attention for Biological Cultures

Computational Modeling of Alzheimer's disease with Real Patients’ Neuron Cell

Pioneering Reinforcement Learning in Mesh Generation

AI-Driven Shape Memory Material Design for 3D Printing

Design and 3D Printing of Suspended Edible Material