Shikun Liu
Senior Research Scientist at Meta
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolShikun Liu is a Senior Research Scientist with 11 years of experience advancing multi-modal and world-modeling research across leading AI labs including DeepMind, Meta, and NVIDIA. He has led core contributions to spatial and video foundation models (Kaleido, MovieGen, CreativeGen) and improved practical training workflows, exemplified by commits to the widely referenced Prismer repository for multi-task vision-language experts. His work spans embodied AGI, camera and motion control for generative video, and efficient post-training techniques, bridging deep research with production-focused engineering. A PhD-trained machine learning researcher from Imperial College London, he blends rigorous academic foundations with hands-on model engineering and checkpoint management. Notably, he has a track record of shipping reproducible code and tooling for complex multi-expert training pipelines, and holds patents from earlier medical imaging work. Based in London, he balances cutting-edge model development with pragmatic system improvements that accelerate large-scale experimentation.
11 years of coding experience
2 years of employment as a software developer
Doctor of Philosophy (Ph.D.), Machine Learning, Doctor of Philosophy (Ph.D.), Machine Learning at Imperial College LondonBachelor of Science (B.S.), Mathematics (with Honors), Electrical Engineering, Bachelor of Science (B.S.), Mathematics (with Honors), Electrical Engineering at Penn State University
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (8)
pytorch10
machine-learning10
trainings10
python10
modeling10
image-annotation9
image-text9
yaml7
Programming languages (5)
Github contributions (5)
NVlabs/prismer
Mar 2023 - Mar 2023
The implementation of "Prismer: A Vision-Language Model with Multi-Task Experts".
Role in this project:
ML Engineer
ML EngineerContributions:22 commits, 4 PRs, 24 pushes in 13 days
Contributions summary:Shikun made several commits related to model training and checkpoint management. They added features to load pre-trained checkpoints for resuming training in multiple training scripts, including `train_classification.py`, `train_pretrain.py`, `train_caption.py`, and `train_vqa.py`. The user also updated the automated download script, `download_checkpoints.py`, to include new models and expert weights. Furthermore, they fixed import issues in several training and demo scripts.
image-captioninglanguage-modelmulti-modal-learningmulti-task-learningvision-and-language
lorenmt/maxl
Dec 2018 - Dec 2021
Contributions:35 commits, 21 pushes, 15 comments in 3 years
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.