Reliability, at depth.

I work where production infrastructure meets machine learning: Kubernetes, Terraform, observability, and networking on one side; a graduate research background and hands-on MLOps on the other. That combination is what it takes to run GPU platforms and training pipelines that researchers can rely on — and to retire the toil that keeps them from scaling.

• CKA certified
• Kaggle Competitions Expert
• OSS: Ray · DeepEval · Loki Helm chart
• K8s / Terraform / GPU infra

1. ▾ 0 m Apr 2024 — present

   Site Reliability Engineer III · F5

   • Shipped an AppStack feature that installs NVIDIA drivers and CUDA for modern GPUs across managed Kubernetes clusters.
   • Enabled deployment on KVM-class environments with L4 GPU passthrough via AppStack VMs.
   • Took over a stalled Loki logging-platform migration and led it to completion — on-premise to EKS with an S3 backend, eliminating hardware-capacity toil and enabling on-demand scale-out.
   • Retired the in-house templating system behind the Loki deployment in favor of the official Helm chart, aligning operations with community standards.

2. ▾ 150 m 2024

   MLOps Engineer · Cluster (freelance)

   • Tuned training-data load performance on an HPC system for ML using striping — accelerating training by ~30%.
   • Built a distributed training pipeline with Ray and Weights & Biases on AWS.
   • Drove an LLM chatbot project — evaluating models, methods, and vector DBs — then handed the system to SWEs.

3. ▾ 300 m Aug 2022 — Apr 2024

   DevOps Engineer · Binance

   • Led development of a GitOps CI/CD pipeline for Terraform from scratch, integrating Atlantis and Policy-as-Code with OPA.
   • Directed a VPN migration across multiple AWS accounts using Transit Gateway, VPC endpoints, and Client VPN.
   • Revived legacy petabyte-scale Hive data pipelines — automated, visualized, and documented them with backfill and retry in a semi-idempotent design.

4. ▾ 450 m Jul 2021 — Jul 2022

   Co-Founder / Tech Lead · MaskTap

   demo video ↗
   • Designed and built an auto-anonymization product from scratch around an extensible API.
   • Developed a framework that automatically masks PII in compliance with GDPR and APPI — original enough to be considered for a patent application.

5. ▾ 600 m Nov 2020 — Jul 2021

   Machine Learning Engineer · Research Center for Advanced Science and Technology, The University of Tokyo (freelance)

   arXiv paper ↗
   • Created a computer-vision segmentation model end to end, starting from planning what data to collect and how much.
   • Diagnosed an unbalanced, undersized dataset in the first PoC and delivered a model that generalizes — the project resulted in an arXiv publication.

6. ▾ 750 m Apr 2020 — Aug 2022

   Site Reliability Engineer · SoftBank

   • Built and operated a production Kubernetes ML-inference platform, with scheduled autoscaling that absorbed traffic surges during large events.
   • Designed end-user APIs on the AI platform using Kubernetes CRDs while preserving compatibility for existing users.
   • Shipped a serverless Vue web app designed for easy maintenance and scale.

7. ▾ 900 m Jul 2017 — 2022

   ML Engineer / Tech Lead · BEENOS (part-time → freelance)

   • Led a team of five as tech lead: Python, Node, AWS, Terraform, GitHub Actions, Kubernetes.
   • Built an ML system for importance discrimination end to end — data collection through deployment.
   • Replaced a legacy monolith crawler with a Pub/Sub architecture for reliability.