I am a Principal Platform Engineer and Cloud Architect based in Berlin. I build the infrastructure that makes production AI systems reliable, and the internal platforms that let product engineers ship without waiting on operations.

Currently: Principal Engineer for ML, Data and Platform Engineering at Roam. Founder of Shipmoor, and Founder & Principal Consultant at Drizzle Systems. Previously Deputy Head of Cloud and SRE at Spryker.

Open to Principal/Staff IC and hands-on engineering leadership roles in AI platforms, ML infrastructure, or Platform Engineering. Also available for consulting, freelance, and advisory engagements (architecture reviews, platform builds, or fractional tech-leadership) If you're building an AI-native platform and need someone who's done it before, let's talk: work.aymen.segni@gmail.com

• Agentic AI infrastructure. Serving layers for autonomous agents and reasoning loops on Kubernetes, built with vLLM, KServe, and KubeRay. I also build MCP servers that let agents operate cloud accounts, and I decide what that surface is allowed to do.
• Platform engineering. Internal developer platforms with self-service provisioning and golden paths. I have established them from greenfield at Spryker, AUTO1, and ING.
• Reliability and cost. SLO programs, incident practice, and observability rebuilds. At Spryker, I cut platform cost by 1.5 million dollars a year, and the result was then sold to five enterprise customers as a product.

I founded Shipmoor, a verification platform for code written by AI agents.

Blocking verdicts come only from deterministic probes. Model inference never decides a verdict. Evidence ships as in-toto attestations, and two runs are compared byte for byte rather than trusting a rerun. The command line tool, a VS Code extension, and an agent harness are shipped. An AI lab runs it daily in continuous integration.

Member of Kubernetes SIGs, the Gateway API Inference Extension working group, and the Knative Serving working group.

Acknowledged by the Microsoft Security Response Center for co-disclosing an AKS privilege-escalation misconfiguration in 2020, fixed and backported across releases.

Lightning talk, Berlin 2025: self-hosting DeepSeek-R1 on AWS EKS with vLLM for inference, KubeRay for distributed scaling, and Terraform with Karpenter for infrastructure.

• 2026 to now, Roam AI. Own the platform and data architecture end to end. This includes hybrid GPU-accelerated Kubernetes on bare metal and a customer-hosted deployment built for data residency.
• 2022 to 2025, Spryker. Designed AI infrastructure for multi-GPU, multi-agent, and multi-LLM workloads. Rebuilt the observability stack on OpenTelemetry, Prometheus, and Grafana with Loki and Tempo, saving 1.5 million dollars a year. Operated more than 300 customer AWS environments across more than 3,000 services. Delivered an internal developer platform on Terraform, GitHub Actions, Atlantis, and Argo CD.
• 2022, AUTO1. Led five staff SREs. Rolled out an SLO operating model that hundreds of service teams adopted, across more than 5,000 microservices.
• 2020 to 2022, ING and Lendico. Led the cloud-native transformation on Azure AKS. Built in-house infrastructure tooling and GitOps for a regulated banking subsidiary.
• 2019 to 2020, Deutsche Bank and Yunar. Ran about 1000 microservices on Azure AKS with an Istio service mesh and a full SLO practice.

Python

• AWS MCP Gateway
• Pre-commit SOPS Hook
• ZFS File Storage Manager API
• Incident Manager Slack Bot

Kubernetes operators

• Kube Admission Controller
• Hibernates (work in progress)

Terraform

• vLLM EKS Helm Module
• EKS vLLM Production Stack Blueprint
• AWS OpenID Connect for GitHub Actions
• Linkerd2 Helm Module
• GKE and PostgreSQL Terraform

• AI infrastructure: vLLM, KServe, KubeRay, MLflow, LangGraph, MCP and FastMCP, Vertex AI, GPU scheduling, NVIDIA drivers on bare metal, fine-tuning with LoRA and quantization
• Cloud: AWS, Google Cloud, Azure, Hetzner, OpenStack
• Kubernetes: EKS, GKE, AKS, K3S, bare metal, operators and CRDs with Kubebuilder, Karpenter, KEDA, Istio, Linkerd
• Platform engineering: Backstage, Argo CD, Atlantis, internal developer platforms, golden paths
• Infrastructure as code: Terraform, Terragrunt, Ansible, Kustomize, Helm, GitHub Actions
• Observability and SRE: OpenTelemetry, Prometheus, Grafana with Loki, Tempo, and Mimir, Datadog, SLOs, error budgets, incident management
• Data: PostgreSQL, Redis, Elasticsearch, ClickHouse, Kafka, Ceph
• Security and compliance: SOC 2, ISO 27001, OPA Gatekeeper, Kyverno, HashiCorp Vault, supply chain security
• Languages: Python, Go, TypeScript

Google Cloud Professional Cloud Architect (2019). Google Cloud Associate Cloud Engineer (2018).

• Website: aymensegni.com and drizzle.systems
• LinkedIn: linkedin.com/in/aymen-segni
• X: @axsegni