starflinger.eu

What It Is

Starflinger is an independent AI infrastructure and self-hosting consultancy aimed at Austrian businesses. Its focus is helping clients run AI and critical services on their own hardware instead of relying on public cloud platforms. It is positioned not as a single SaaS tool, but as an implementation consultancy around GPU servers, Proxmox clusters, LLM inference, RAG, developer workflows, and critical IT services.

Core Capabilities

On the AI side, the page explicitly mentions vLLM, SGLang, LiteLLM, multi-GPU setups, model routing, autoscaling, and token throughput monitoring, suggesting coverage of the infrastructure layer from inference serving to operational monitoring. For models, it supports local hosting, routing, and documentation of open-source models such as Llama, Mistral, Qwen, and DeepSeek, with an emphasis on auditability and optional air-gapped deployments. Its RAG capabilities include embedding pipelines, pgvector, Qdrant, BGE/E5, re-ranking, and eval loops, making it suitable for retrieval across internal documents, codebases, and support tickets.

Pricing and Integrations

The page does not disclose pricing, free trials, service packages, or payment methods, so it is not possible to assess the budget threshold. In terms of integrations, beyond AI inference and RAG components, it also covers Claude Code, Codex, Proxmox, Vault, Stalwart, Wazuh, CI Runner, monitoring, email, and backups, pointing toward the build-out of a complete self-hosted technology stack. Consulting is available in German and English, with no mention of Chinese-language support.

Pros and Cons

The main strength is its very clear focus on data sovereignty: no cloud lock-in, reduced data outflow, and maintaining control on owned hardware, along with references to DSGVO/GDPR and NISG 2026 compliance direction. The consultant claims to run a similar 4-node Proxmox cluster and a dual RTX PRO 6000 GPU server setup, suggesting a practical, hands-on orientation. The limitations are mainly around transparency: there are no customer case studies, SLA details, pricing, delivery timelines, or team-size information, nor are there quantified performance or quality benchmark results.

Who It’s For and Access from China

Starflinger is best suited to medium and large enterprises in Austria and the wider German-speaking region, regulated industries, IT teams looking to privately deploy LLM/RAG systems, and organizations planning to move away from SaaS lock-in. Access and payment conditions for users in China are unknown. Given its local consulting nature and German/English language service, Chinese companies needing similar solutions would typically consider local private-deployment LLM providers, Alibaba Cloud PAI, Baidu Qianfan, Volcano Ark, or building an open-source stack with Ollama, vLLM, Qdrant, and similar tools.