About

Self Introduction

• Yuki Iwai (he/him)
• Tokyo, Japan (UTC+9)
• Software Engineer and Part-time OSS developer at CyberAgent, Inc.

Technical Interest

• AutoML
• Distributed Trainig
• Batch Workload
• Storage
• Kubernetes

OSS Activity

I’m focuced on developing the Kubernetes-based Distributed Systems for AutoML, Distributed Training and Batch Workloads.

• Member of Kubernetes Organization
  • kubernetes-sigs/kueue (SIG Scheduling / WG Batch) active contributor
• Member of Kubeflow Organization
  • kubeflow/katib (WG AutoML) approver
  • kubeflow/training-operator (WG Training) reviewer
  • kubeflow/mpi-operator (WG Training) active contributor
• Member of Kserve Organization
  • kserve/kserve contributor

Experience

• 2022/04 - current: Software Engineer (Private Cloud) and Part-time OSS developer at CyberAgent, Inc.
  • Development of the on-prem Kubernetes as a Service (KaaS).
    • Security Policies for many Kubernetes Clusters using open-policy-agent/gatekeeper.
  • Development of the on-prem Kubernetes-based Machine Learning Platform.
    • Hyperparameter Tuning System
    • Job Systems for trainig ML models and predicting target values (inference)
    • Distributed Training System (RDMA/GPU)
    • Serving system for ML models
    • Full managed interactive development environment (JupyterLab or Jupyter Notebook)
• 2021/07 - 2021/11: Part-time Infra/Software Engineer (Private Cloud) at CyberAgent, Inc.
  • Development of the on-prem Kubernetes as a Service (KaaS).
    • Security Policies for many Kubernetes Clusters using open-policy-agent/gatekeeper.
  • Development of the on-prem Kubernetes-based Machine Learning Platform.
    • Serving system for ML models
    • Job Systems for trainig ML models

Internship

• 2020/10/01 - 2020/10/31: CyberAgent, Inc.
  • Survey of Kubeflow features
  • Survey of NVIDIA DGX A100 performance and features
  • Blog: https://developers.cyberagent.co.jp/blog/archives/27764/ (Japanese)
• 2020/09/02 - 2020/09/15: Yahoo! Japan
  • Development of monitoring infrastructure for on-prem Kubernetes as a Service (KaaS)
• 2020/08/03 - 2020/08/14: Cybozu
  • Development of Rook on upstream
• 2020/07/20 - 2020/07/28: F@N Communications
  • Survey of Vitess features
  • Blog: https://n.fancs.tech/blog/beginnerofvitess/ (Japanese)

Education

• 2020/04 - 2022/03: Electronic Engineering Major, Graduate School of Science and Engineering, Kindai University
  • Master of Engineering (Computer Science)
• 2016/04 - 2020/03: Department of Informatics, Faculty of Science and Engineering, Kindai University
  • Bachelor of Engineering (Computer Science)

Talks

• Japan Rook Meetup #3 ‘ML環境でのRook/Ceph’ (Japanese)

Publications

• 入門Kueue〜KubernetesのBatchワークロード最前線〜 (Japanese)