Zhaoqi L
AI Research Scientist at Meta
United States
Summary
🤩
Rockstar🎓
Top SchoolZhaoqi L is an AI research scientist with nine years of experience bridging physics, machine learning, and production-scale perception systems, currently focusing on model self-improvement at Meta. A Princeton Ph.D. candidate in Physics with a BS in Engineering Physics from Cornell, he has led multimodal foundation-model pretraining and AutoML efforts at Waymo and contributed core research and engineering to Gemini projects at DeepMind. His open-source work includes notable contributions to KerasCV, where he implemented advanced 3D point-cloud and bounding-box augmentation techniques that improve robustness for real-world perception tasks. Comfortable moving between deep research and product-facing engineering, he has repeatedly shipped agentic reasoning and post-training improvements for large models. Notably, his background in physics informs a principled approach to modeling uncertainty and data augmentation that yields practical gains in safety-critical perception stacks.
9 years of coding experience
1 year of employment as a software developer
BS, Engineering Physics/Applied Physics, BS, Engineering Physics/Applied Physics at Cornell UniversityDoctor of Philosophy, Physics, Doctor of Philosophy, Physics at Princeton University
Github Skills (6)
keras10
computer-vision10
machine-learning10
bounding-box10
tensorflow10
data-augmentation10
Programming languages (1)
Github contributions (5)
keras-team/keras-cv
Nov 2022 - Jan 2023
Industry-strength Computer Vision workflows with Keras
Role in this project:
ML Engineer
ML EngineerContributions:131 reviews, 17 commits, 17 PRs in 1 month
Contributions summary:Zhaoqi primarily contributed to the development of 3D computer vision data augmentation techniques within the KerasCV framework. Their work involved creating and modifying preprocessing layers, including those for global random rotation, scaling, flipping, and dropping points/boxes. The user's commits demonstrate a strong focus on implementing and refining various data augmentation strategies to improve model training for 3D perception tasks, specifically related to point clouds and bounding boxes. The user also addressed a bug in grouping points by bounding boxes.
visiondeep-learningcomputer-visionindustrystrength
lengzq/lengzhaoqi
Jan 2018 - Feb 2018
Contributions:9 pushes, 1 branch in 22 days
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.