Agustin – Ansible, Linux, Kubernetes

FirmenOnline is Allianz’s B2B platform for the German corporate life-insurance market, used by business customers to manage company life-insurance products. It runs about 40 production services built by six application teams (~60 developers) on a centrally managed cloud platform, where any outage directly affects business clients.

• Own reliability for ~40 production services across six application teams (~60 developers) on centrally managed AWS EKS with Istio/Envoy, defining internal SLIs/SLOs and leading P0/P1 incident response;
• Designed and delivered “Dynamic Stage”, a Kubernetes operator and CI suite that lets teams run several versions of a service side by side in one namespace with Istio conditional routing and TTL-based cleanup. Business analysts pick which version to test on a self-service page. Took it from research and stakeholder alignment through several prototypes to full production use;
• Consolidated 30+ service Helm charts into three reusable deployment primitives. Replaced a complex Helm library with a simpler base chart so product teams can read and maintain their own deployments, and standardized CI/CD (GitHub Actions, Jenkins) across 30+ applications;
• Run production deployments and promotions through Argo CD (GitOps) alongside Helm and GitHub Actions/Jenkins;
• Moved observability to Dynatrace Configuration-as-Code (dashboards, custom alerts, synthetic monitors), and turn P0/P1 investigations into runbooks, RCAs and systemic fixes within ServiceNow processes;
• Built a pilot agentic incident-investigation platform (LangGraph, Pydantic, Claude on AWS Bedrock). It reacts to Dynatrace problems and posts evidence-based RCA and remediation summaries to Microsoft Teams. Instrumented it with OpenTelemetry and Langfuse (latency, token cost, failures, tool calls) and led its security audit, which closed a tool-permission gap before go-live;
• Drive AI-assisted engineering adoption with Claude Code skills, agents, MCP integrations and GitHub Actions code-review automation.

Shared digital infrastructure for CaixaBank, one of Spain’s largest banks. It hosts hundreds of production services across several cloud environments and serves the bank’s application teams, with high requirements for availability and cost control.

• Built a monthly QoS screening pipeline over production Prometheus metrics in Thanos. It ranks hundreds of services by degraded golden signals (latency, error rates, saturation, disk pressure) and feeds an evidence-based remediation backlog for application teams;
• Administered production Kubernetes platforms on IBM Cloud and OpenShift with Istio service mesh and Argo CD GitOps deployments;
• Designed observability solutions by unifying telemetry sources into shared dashboards and alerting, and advised teams on instrumentation and signal quality;
• Led GCP cost optimization through rightsizing, scaling, and query and storage improvements, and developed reusable Terraform modules;
• Designed and built an AI-driven code-analysis platform (Python, Vertex AI/Gemini) that surfaces performance and cost-optimization findings for application teams and automates their analysis workflows;
• Led research into multi-agent systems and codebase indexing, turning new AI capabilities into concrete engineering use cases.

A large-scale hosting infrastructure behind the web-hosting products and internal platforms of one of Europe’s largest hosting and cloud providers. It spans thousands of servers and more than 100k nodes across six company-owned datacenters, where any capacity or availability issue directly affects hosted customer services.

• Automated region migrations, OS patching and failovers with Ansible, cutting execution from days to minutes and reducing downtime by more than 50%;
• Co-engineered the move from legacy monitoring to a telemetry platform covering thousands of servers across six datacenters: custom Python agents on LXC and bare metal, Redis-queued aggregation, InfluxDB/PostgreSQL storage and Grafana alerting;
• Reduced high-load incidents by about 30% and eliminated CPU contention across 100k+ nodes through utilization and capacity analysis;
• Operated bare-metal Kubernetes (Rancher RKE, MetalLB) and KubeVirt-based infrastructure with Istio and Argo CD GitOps, and owned production Longhorn storage (CSI, StorageClasses, PV/PVC);
• Ran the self-hosted Graylog/ELK stack, evaluated Loki as a replacement and kept Graylog after a technical assessment;
• Provided L3 support and on-call coverage for critical Linux incidents, and automated malware detection and remediation using IOC and behavioural analysis.

Advanced technical support for Google Cloud’s largest enterprise customers running production workloads on Google’s managed container platform. The team worked directly with Google’s own SRE and engineering teams to resolve critical availability and performance issues and product bugs.

• Diagnosed complex availability, performance, networking and workload issues on GKE for top Google Cloud enterprise customers;
• Worked with Google SREs and SWEs on deep investigations and confirmed product bugs, managing escalations through Salesforce;
• Advised customers on Kubernetes observability and production troubleshooting practices;
• Guided enterprise teams on cloud-native architectures and production operating practices.