Kazuki Odo

About

Experience

1. Oct 2025 — PresentTokyo

   Staff Software Engineer

   Turing, Inc.

   1. Infrastructure teamJul 2026 — Present

      • ▸Build and operate Slurm-based GPU training clusters — 1,000+ GPUs across on-prem, a domestic GPU cloud, AWS, and Google Cloud — with a small infrastructure team supporting fully autonomous driving development.
      • ▸Extend IaC and CI/CD (Terraform, Terragrunt, GitHub Actions) from the clouds to existing environments, splitting node configuration across boot images (Packer), startup scripts, and Ansible, and rolling out changes by draining nodes first to limit impact on training jobs.
      • ▸Worked with AWS to isolate FSx for Lustre I/O stalls on SageMaker HyperPod and cut over to a rebuilt cluster on custom AMIs; completed a blue-green migration of the Google Cloud cluster's Managed Lustre to a larger instance.
      • ▸Make room for improvement work with observability as code (Datadog), runbooks built from incident records, and AI-assisted first-pass investigation — while operators still assess impact on shared infrastructure and decide what to adopt.
      • ▸Share platform knowledge through user guides written for both people and AI agents, and internal infrastructure study sessions.

   2. MLOps teamOct 2025 — Jun 2026

      • ▸Built and operated cloud GPU clusters for autonomous-driving model training: a SageMaker HyperPod + Slurm + FSx for Lustre cluster on AWS (AWS CDK, Terraform), and a 240-GPU (NVIDIA H200) Slurm cluster on Google Cloud stood up in two weeks (Cluster Toolkit, Terraform / Terragrunt, Packer, Ansible).
      • ▸Designed and developed a dataset-management system for large-scale training data, automating dataset download and cleanup on cluster storage.
      • ▸Built the initial prototype for streaming video training data directly from S3 — parallel-fetch benchmarks and the initial pipeline design — before moving to the infrastructure team; other MLOps engineers completed the system and published it on the Turing Tech Blog.
      • ▸Automated detection and replacement of failed nodes on SageMaker HyperPod.
2. Jun 2024 — Sep 2025Tokyo

   Senior Site Reliability Engineer

   Splunk (Cisco)

   • ▸Maintained and improved Splunk Cloud infrastructure across AWS, GCP, and Azure.
   • ▸Designed the Professional Services delivery menu for provisioning ISMAP-compliant Splunk Cloud instances (Japanese government cloud security program).
   • ▸Automated operational workflows and improved system reliability while concurrently serving on the Fulfillment Engineering team.
   • ▸Supported customer environments and collaborated with account teams, solution engineers, Professional Services, and Technical Support.
   • ▸Contributed to incident response and operational excellence across APAC.
3. Apr 2023 — May 2024Tokyo

   Senior Software Engineer

   Preferred Networks

   • ▸Developed MATLANTIS, an enterprise-scale atomic-level simulation platform.
   • ▸Led multi-GPU, multi-cloud workloads on Kubernetes multi-cluster.
   • ▸Built infrastructure and tooling using Terraform, Kubernetes, and Bazel.
   • ▸Delivered backend systems and data pipelines supporting product and business use cases.
4. Jun 2021 — Mar 2023Tokyo

   Customer Engineer — Google Cloud (Gaming, Media & Entertainment)

   Google Cloud Japan

   • ▸Advised enterprise customers on cloud architecture, migration, and modernization.
   • ▸Designed and prototyped solutions using GCP across application and ML domains.
   • ▸Resolved 200+ technical issues across customer engagements.
   • ▸Led technical strategy for customer transformation and platform adoption.
5. Oct 2017 — May 2021Tokyo

   Security / Software / Site Reliability Engineer

   Recruit

   • ▸Built and operated backend systems for large-scale services using Spring Boot and cloud infrastructure.
   • ▸Led migration from private cloud (VMware) to AWS-based cloud-native architecture, improving scalability and operational efficiency.
   • ▸Improved system reliability through monitoring, performance tuning, and operational enhancements.
   • ▸Analyzed logs and developed detection systems for account takeover and fraudulent behavior using Splunk.
   • ▸Led cloud adoption strategy across AWS and GCP, focusing on security architecture and enterprise-scale rollout.
6. Apr 2013 — Sep 2017Tokyo

   Network & Security Research Engineer

   NTT Communications (NTT DOCOMO BUSINESS)

   • ▸Contributed to network security and traffic-control platforms, including work reflected in public patents and publications.
   • ▸Designed and built two internal on-premises platforms from scratch: a VMware-based IaaS and a Kubernetes-based container platform — both adopted as long-running production infrastructure.
   • ▸Developed DDoS protection and traffic control systems.
   • ▸Researched network security technologies including BGP Flowspec.
   • ▸Built distributed systems for traffic analysis and monitoring.

Skills

Side Projects

Selected side projects done alongside my full-time roles.

Open Source

Selected OSS contributions.

Education & Certifications

Contact

Open to conversations about ML platforms, Infrastructure, SRE and Software development.

LinkedIn preferred for first contact.