How to become an AI infrastructure engineer?
A backend engineer working on a GenAI platform with 48x H200 GPUs, Kubernetes, and vLLM seeks advice on transitioning from an SRE-focused role (deploying and serving LLMs) into a proper software engineer specializing in AI infrastructure. They ask for recommendations on distributed systems, difficult-to-acquire skills, and resources outside of large-scale companies.
Background
- This is a Hacker News "Ask HN" post: a community Q&A thread where a working engineer asks peers for career and learning advice. The author already has a job building AI serving infrastructure (working with NVIDIA H200 GPUs, Kubernetes, and vLLM) but feels their role is more operations/site-reliability (SRE) than true software engineering, and wants to bridge that gap.
- Key terms: **vLLM** is a popular open-source inference engine for large language models; **H200** is NVIDIA's high-end GPU for AI workloads; **inference infrastructure** is the systems that run trained models in production (as opposed to training them).
- Why it matters: As AI deployment becomes mainstream, "AI infrastructure engineer" is an emerging, ill-defined role — part DevOps, part systems engineering, part ML-adjacent. The post reflects a common anxiety among practitioners: are they building genuine, transferable engineering skills, or just operating the latest trendy tools?