Babak Zarrinbal

Cloud & Platform Engineer

Kubernetes · GitOps · LLM Infrastructure — Düsseldorf, Germany

14+ years in software, 10+ years leading engineering teams, now focused on Kubernetes-native cloud platforms and LLM infrastructure. Currently building multi-cloud (Azure/GCP) platforms and agentic AI systems at Douglas. Active upstream contributor to kgateway (CNCF Envoy-based API/AI Gateway).

Open Source — kgateway (CNCF)

merged PR #14231 — fixed XListenerSet status validation
merged PR #13592 — exposed headersToClient in ext-authz responses
root-caused Issue #14363 — readiness race serving partial xDS config on a 3,500+ cluster production gateway; validated with a 1M-request load test, fix merged as PR #14380

Experience

Douglas

Mar 2025 – Present

Cloud Ops Engineer

  • Own multi-cloud IaC (Azure DevOps + Terraform, GCP + Terraform) for production and staging Kubernetes clusters
  • Implemented kgateway from scratch and migrated all production routing onto it
  • Built standalone agentic applications on Claude and Gemini (via Copilot) with RAG and memory management, under strict internal data-handling constraints
  • Built containerized, fully dynamic Azure self-hosted CI/CD agents, autoscaled with Karpenter and KEDA ScaledJob
  • Manage a team of DevOps engineers

managbl.ai (acquired by casavi, Sep 2025)

Apr 2024 – Mar 2025

Head of DevOps / Development Manager

  • Berlin PropTech automating phone accessibility for property managers via AI; tech now serves 150+ customers via casavi
  • Deployed and served Hugging Face models with vLLM; fine-tuned models on masked/anonymized user data for intent classification
  • Built a RAG pipeline to retrieve contract terms, restrictions, and provider info for real-time cross-checking
  • Migrated infrastructure to Kubernetes with spot instances to cut compute cost while scaling

Retraced

Feb 2022 – Mar 2024

Team Lead / Technical Manager

  • German blockchain-based supply-chain traceability platform for fashion (2,000+ brands and suppliers)
  • Led architecture overhaul and rewrote core services into dockerized components
  • Shipped an AI-assisted document management feature (OpenAI, Anthropic APIs) and a Corrective/Preventive Actions (CA/PA) feature on a dynamic, open-schema data layer over Oracle DB

Tecnotree

Sep 2016 – Feb 2022

Technical Manager

  • Finnish telecom vendor; delivered DCBS (Digital Convergent Billing System), used by operators including Ooredoo
  • Managed ~25 engineers across 5 rotating teams, a QC team, and a 4-person DevOps team on Rancher-based Kubernetes

Technical Skills

AI / LLM

vLLM, Hugging Face Transformers, fine-tuning, RAG, agentic apps, memory management, OpenAI, Anthropic (Claude), Gemini

Platform

Kubernetes, ArgoCD, Karpenter, kgateway/Envoy, cert-manager, Kyverno, KEDA, Rancher, Docker, Terraform

Cloud

GCP, Azure, AWS (ECS, EC2, Lambda, EKS, S3, DynamoDB), Oracle Cloud

Languages

JavaScript/TypeScript (Node.js, NestJS, React, Vue), Python, Go, Java (Spring Boot), Bash

Data & Messaging

Oracle DB, MongoDB, Cassandra, PostgreSQL, Kafka, RabbitMQ, Redis

Security & Observability

CrowdStrike Falcon, secret rotation, Datadog, Grafana