Bonhun Koo
Software Engineer at 구글
Seoul, South Korea
Summary
🤩
Rockstar🎓
Top SchoolBonhun Koo is a software engineer with four years focused on productionizing ML for large-scale, real-time computer vision systems, currently working on MTMC tracking at Google. He designs end-to-end architectures that handle 30+ camera streams, develops detection and re-identification models in Python, and ports high-performance components to C++/TensorRT for inference at scale. Previously he led MTMC and medical image anonymization efforts at Deeping Source and added INT8/mixed-precision quantization and dataset tooling contributions to prominent open-source projects like OpenVINO and Datumaro. His master’s research produced an award-winning news recommender published at PAKDD20, reflecting a blend of applied research and systems engineering. Based in Seoul, he pairs low-level performance optimization experience from Samsung and Intel with a practical knack for turning research prototypes into robust, deployable pipelines.
3 years of coding experience
11 years of employment as a software developer
Bachelor's degree, Computer Science & Engineering, Bachelor's degree, Computer Science & Engineering at 서울대학교 (Seoul National University)Master's degree, Computer Science & Engineering, Master's degree, Computer Science & Engineering at Seoul National University
Github Skills (15)
quantization10
computer-vision10
inference10
deep-learning10
coco10
python10
ai10
openvino10
optimization10
datasets10
deeplearning-ai9
jupyter-notebook9
mlops8
tensorflow5
pytorch5
Programming languages (2)
Github contributions (5)
open-edge-platform/datumaro
Oct 2022 - Dec 2022
Dataset Management Framework, a Python library and a CLI tool to build, analyze and manage Computer Vision datasets.
Role in this project:
ML Engineer
ML EngineerContributions:107 reviews, 8 commits, 18 PRs in 1 month
Contributions summary:Bonhun focused on enhancing the dataset management capabilities of the `datumaro` library, specifically related to computer vision datasets. Their contributions include adding a Jupyter notebook to merge heterogeneous datasets for object detection and implementing functionalities for hierarchical labeling and supporting export as CVAT video format. Furthermore, the user worked on enabling bounding box annotations with segment annotations for the COCO format.
pythoncomputer-vision-datasetsclipascal-voccoco
openvinotoolkit/openvino
Jun 2022 - Sep 2022
OpenVINO™ is an open-source toolkit for optimizing and deploying AI inference
Role in this project:ML Engineer
ML EngineerContributions:25 reviews, 6 commits, 12 PRs in 3 months
Contributions summary:Bonhun contributed to the OpenVINO toolkit, focusing on optimizing AI inference. Their work involved adding INT8 quantization support to the GNA speech sample and preventing potential overflow issues within the POT framework's quantization algorithms. The user also implemented output quantization schemes and updated IndexSampler usage within the optimizer, enhancing the quantization capabilities of the OpenVINO toolkit. Furthermore, the user contributed to Accuracy Aware Quantization by adding mixed-precision support for GNA and providing sanity tests.
inference-enginepytorchmodel-optimizerdeep-learninggpu
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.