Xiaopeng Li

Xiaopeng Li
AI researcher with ten years of experience. Primarily interested in Large Language Models training, Reinforcement Learning from Human Feedback, Natural Language/Code Generation, LLM Agent, Retrieval Augmented Generation, etc.

About me

I am currently a Senior Applied Scientist in Amazon AWS. I completed my Ph.D. at the Hong Kong University of Science and Technology in 2019. I have worked primarily on large language models and generative AI. As a science lead, I have initiated and successfully launched two products: Amazon CodeWhisperer and Amazon Q in IDE. I am passionate about doing original and impactful research in AI and ML, and enjoy working with smart people on exciting projects. I like to create proof-of-concepts and new products with the latest technologies.

I think learning, exploration and creation are life-time endeavours.

My experience

Research

Project I worked on

Training LLMs to Better Self-Debug and Explain Code - preprint 2024

Nan Jiang, Xiaopeng Li et al.

Project I worked on

Multi-lingual Evaluation of code-generation model - ICLR 2023 (spotlight)

Ben Athiwaratkun, Sanjay Krishna Gouda, Zijian Wang, Xiaopeng Li, et al.

Project I worked on

CONTRACLM: Contrastive Learning For Causal Language Model - ACL 2023

Nihal Jain, Dejiao Zhang, Wasi Uddin Ahmad, Zijian Wang, Feng Nan, Xiaopeng Li, et al.

Project I worked on

Exploring Continual Learning for Code Generation Models - ACL 2023

Prateek Yadav, Qing Sun, Hantian Ding, Xiaopeng Li, et al.

Project I worked on

Not All Attention Is Needed: Gated Attention Network for Sequence Data - AAAI 2020

Lanqing Xue, Xiaopeng Li, Nevin L. Zhang

Project I worked on

Learning Latent Superstructures in Variational Autoencoders for Deep Multidimensional Clustering - ICLR 2019

Xiaopeng Li et al.

Project I worked on

Collaborative Variational Autoencoder for Recommender Systems - KDD 2017

Xiaopeng Li and James She.

Demo

Project I worked on

Code AutoCompletion with LLM

I pretrained a code GPT model in Aug 2020 for code generation, and created a interactive demo for code completion in IDE. Very primitive, but it was 2020 way before LLM surge.

Project I worked on

Chat LLM

I created the first chat LLM in the code generation domain in Feb 2023, and created a interactive chat demo. Very primitive, but it was right after ChatGPT appears and before any open source chat model release.