求建议：后端工程师 → 基础设施工程师：如何实现转型？
4 分•作者: gokuljs•3 个月前
我做了大约四到五年的后端工程师，主要是在初创公司。大约三个月以来，我一直在阅读和构建一些小东西，但我不确定自己是在进步还是原地打转。而且，我周围也没有这方面的人，所以大部分时间都是自己摸索，包括使用 GPT 和 Claude 这样的工具，但我仍然感到困惑。
我的工作包括 API、一些实时系统（如 WebRTC 和流媒体），以及调试生产问题，例如延迟、缓冲和可靠性。我使用过 AWS、Docker、Redis 和类似的工具。
我正试图转向更多面向系统的工作，例如基础设施、分布式系统，或者像推理管道和数据流这样的 AI 基础设施。
问题是我感到停滞不前，而且目标分散。我一直在基础设施、SRE 或平台工作，以及 AI 基础设施之间跳来跳去，而且我不太清楚从我目前的位置开始，实际的路径是什么样的。
一些问题：
* 人们实际上是如何从后端或产品工作转向基础设施或系统角色的？
* 是应该尽早选择一个方向，还是先广泛探索？
* 如何选择一个可以随着时间推移而真正积累，而不是不断自我怀疑的主题或方向？
* 什么样的项目真正表明了准备就绪？
* 实际上，在没有基础设施头衔的情况下，如何开始获得面试机会？
非常感谢那些已经完成这种转变的人的建议。
谢谢
查看原文
I’ve been a backend-heavy engineer for about 4 to 5 years, mostly in startups. For about 3 months I’ve been reading and building small things, but I’m not sure if I’m progressing or just spinning. I also don’t really have people around me in these areas, so I’ve mostly been trying to figure this out on my own, including using tools like GPT and Claude, but I still feel unclear.<p>My work includes APIs, some real-time systems like WebRTC and streaming, and debugging production issues such as latency, buffering, and reliability. I’ve used AWS, Docker, Redis, and similar tools.<p>I’m trying to move toward more systems-oriented work, such as infrastructure, distributed systems, or AI infrastructure like inference pipelines and data flow.<p>The problem is I feel stuck and scattered. I keep jumping between directions such as infrastructure, SRE or platform work, and AI infrastructure, and I don’t have a clear sense of what the actual path looks like from where I am.<p>Some questions:<p>How do people actually move from backend or product work into infrastructure or systems roles?<p>Is it better to pick one direction early, or explore broadly first?<p>How do you pick a theme or direction that actually compounds over time instead of second-guessing constantly?<p>What kinds of projects really signal readiness?<p>And practically, how do you start getting interviews without already having an infrastructure title?<p>Would really appreciate advice from people who’ve made this transition.<p>Thanks