PK
Principal Platform Engineer · Cogent Labs, Tokyo

Parth
Kakadia

Platform engineer in Tokyo. I build the platforms ML teams ship on, and cut what they cost to run.

View experienceGet in touch
01 · Experience

Where I've worked

2024 –Dec 2024 – Present
Tokyo, Japan

Cogent Labs · Principal Platform Engineer

AI document processing
  • Cut cloud costs by 50% in three months with no impact on performance: S3 retention policies, Kubernetes rightsizing, and spot capacity with on-demand fallback.
  • Built cost attribution that calculates the unit cost of every field, page and document processed, and margin tooling leadership uses for pricing.
  • Enabled GPU sharing across ML inference workloads, replacing one-pod-per-node allocation.
  • Built the observability stack from scratch on Prometheus, Thanos, Loki, Tempo and Grafana, fully managed as code.
  • Led the migration from Flux to ArgoCD, giving teams self-service GitOps deployments.
  • Re-architected environments as multi-cluster blue/green EKS on Terraform and Terragrunt; delivered partner connectivity over Direct Connect and PrivateLink.
Kubernetes (EKS)ArgoCDTerraformKarpenterPrometheusGrafanaAWSGCP
2021 –Aug 2021 – Nov 2024
India (remote)

Measured · Software Engineer III, Platform

Advertising data platform
  • Designed a workflow-as-code platform on Apache Airflow whose dynamic DAG generator runs 1,000+ production ETL pipelines.
  • Built a config-driven, multi-tenant ETL platform that cut client onboarding to 24 hours.
  • Ended noisy-neighbour resource starvation with distributed concurrency control and per-client rate limits on Redis.
  • Led a cross-functional team building a FinOps analytics engine leadership used for pricing and renewals.
AirflowPythonRedisSpring BootAWS
2021Feb – Jul 2021
Hyderabad, India

Cloud Elements · Integration Engineer

API integration platform, acquired by UiPath
  • Designed an API abstraction layer that unifies several ERP and CRM schemas behind one REST interface, so customers integrate once instead of once per vendor.
  • Automated high-volume ETL and data synchronisation workflows across enterprise cloud systems.
2019 –May 2019 – Feb 2021
Gandhinagar, India

Tata Consultancy Services · Assistant System Engineer

Cloud migration and production operations
  • Migrated production workloads from Heroku to a multi-account AWS architecture, building an infrastructure-as-code generator that cut migration time by 60%.
  • Maintained 99.9% availability for a payroll system through database tuning, server lifecycle management and disaster recovery drills.
02 · Skills

What I work with

Kubernetes and platform

EKSGKEKarpenterHelmKnativeCrossplane

Infrastructure as code and delivery

TerraformTerragruntArgoCDGitHub ActionsBlue/green

Observability and reliability

PrometheusThanosGrafanaLokiTempoOpenTelemetryPagerDuty

Cloud and cost

AWSGCPKubecostSpot strategyUnit economics

Data and pipelines

KafkaAirflowRedisPostgreSQLRedshift

Languages

PythonGoTypeScriptJavaSQL
03 · About

I've spent seven years on the infrastructure other engineers build on, from a payroll system at TCS to the platform behind AI products in Tokyo.

I like self-service over tickets, infrastructure as code over clicking in consoles, and knowing what everything costs. Lately I've been going deeper into ML infrastructure and model serving.

EducationB.E. Computer Engineering, Gujarat Technological University (2015–2019)
PublicationsIEEE Internet of Things Journal (2020) doi ↗ · Springer LNNS (2023) doi ↗
Based inTokyo, Japan
04 · Contact

Let's talk.

Emailparthkakadia143@gmail.com
LinkedInlinkedin.com/in/parthkakadia
ResumeRequest by email