Mehrshad L.

I help teams running AI and cloud workloads see what their GPUs, models and services are doing, with a managed Grafana stack that runs inside their own cloud account. I'm the founder of Optiop GmbH (Frankfurt, Germany) and the builder of GaaS, a sovereign observability platform for AI infrastructure. GaaS provisions and operates the open-source Grafana LGTM stack (Grafana, Mimir, Loki, Tempo, Alloy) in your AWS, Azure or StackIT account. Your metrics, logs and traces stay in your infrastructure, and nothing of ours sits in the data path. WHAT I CAN DO FOR YOU - Managed Grafana LGTM in your own cloud: provisioned in under an hour, datasources pre-wired across metrics, logs and traces, with upgrades and rollbacks handled for you - GPU monitoring: NVIDIA DCGM dashboards and alerts for utilization, memory, power, XID errors and MIG slices, plus idle-GPU detection - LLM and inference observability: OpenTelemetry GenAI tracing for vLLM, Triton, OpenAI, Azure OpenAI and Anthropic calls, covering latency, time to first token, tokens and errors per model and route - Cost per model and per team: GPU-hours and token spend attributed to the namespace, team or customer that used them - Golden-signal dashboards, SLOs and error budgets for your services, with alerts routed to your on-call tool - EU data sovereignty and EU AI Act record-keeping: long-term logs and traces retained in your own EU account - Kubernetes, Terraform/OpenTofu and CI/CD so dashboards and alerts are versioned as code WHO THIS IS FOR - Teams running GPUs or LLMs in production who want visibility without sending telemetry to a third-party SaaS - European companies with data-residency, GDPR or AI Act requirements - Teams already on Prometheus/Grafana who have outgrown a single instance HOW WE CAN START 1. A short call to understand your cloud, workloads and what you monitor today 2. I provision the GaaS stack into your account. The stack itself is free; you pay your cloud provider for compute and storage 3. You choose the work on top: dashboards, SLOs, GPU/LLM observability, managed upgrades or ongoing support WHY WORK WITH ME - 100% Job Success, Top Rated Plus, 41 jobs and 1,800+ hours on Upwork - Grafana Champion, Kubestronaut (CKA, CKAD, CKS), Prometheus Certified Associate, OpenTelemetry Certified Associate - Background in SRE and infrastructure engineering (Deutsche Börse, Max Planck Institute for Software Systems) - A small German GmbH team, so you get continuity and EU-based accountability, not a lone freelancer Send me a message here on Upwork with what you run and where you run it, and ask for a GaaS walkthrough. I usually reply within a few hours.