Ji Lin
PhD Candidate
Cambridge, England, United Kingdom
Summary
🤩
RockstarJi Lin is a PhD candidate and seasoned machine learning engineer with a decade of experience focused on efficient video understanding and model optimization. Based in Cambridge, UK, Ji has contributed to influential open-source work at MIT’s Han Lab, notably enhancing the ICCV‑2019 Temporal Shift Module with non-local variants, in-place temporal shifts, pretrained models, and an online demo for practical evaluation. Their contributions show a knack for bridging research and production—fixing training issues like weight decay and delivering pretrained assets for Kinetics and Sth‑Sth benchmarks. Ji combines deep academic rigor with hands-on engineering, consistently improving model performance and usability for the broader ML community.
10 years of coding experience
Github Skills (8)
pytorch10
machine-learning10
computer-vision10
video-understanding10
python9
acceleration9
faster-rcnn8
mask-rcnn8
Programming languages (5)
Github contributions (5)
mit-han-lab/temporal-shift-module
Mar 2019 - Aug 2021
[ICCV 2019] TSM: Temporal Shift Module for Efficient Video Understanding
Role in this project:
ML Engineer
ML EngineerContributions:36 commits, 5 PRs, 36 pushes in 2 years 5 months
Contributions summary:Ji contributed to the development and enhancement of the Temporal Shift Module (TSM) for efficient video understanding. They added non-local TSM models trained on Kinetics, an in-place version of the temporal shift, and pretrained models for Kinetics and Sth-Sth datasets. Furthermore, the user fixed weight decay issues and added an online demo for the TSM model. Their work primarily focused on improving model performance and providing practical tools for video analysis.
nvidia-jetson-nanopytorchiccviccv-2019shift
mit-han-lab/mcunet
Jun 2022 - Dec 2022
[NeurIPS 2020] MCUNet: Tiny Deep Learning on IoT Devices; [NeurIPS 2021] MCUNetV2: Memory-Efficient Patch-based Inference for Tiny Deep Learning
Contributions:36 commits, 4 PRs, 13 pushes in 6 months
pytorchmemorypatchneurips-2020neurips-2021
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.