Even startups are rarely doing anything novel with software engineering, computing systems, etc. They're usually stumbling around in the search for a market niche to exploit. It happens that they mostly use software these days to achieve that but it's absolutely a secondary consideration and hardly the most important. If the company manages to survive and gain adoption it's likely that the original code they wrote will get thrown out the window or wrapped up behind layers of glue code.
There's very little work outside of that which isn't dominated by people with decades of experience and expertise who aren't going anywhere because the jobs are few and far between. Compilers? Real-time operating systems? Traffic control systems? Filesystems? Operating systems? Media codecs? Security? Typically the funnel for these jobs come straight out of universities. They don't need to hire generalists, there's a smaller market of job posts for these projects, and they have a well-established base of people who've built up these niche skills and rarely leave jobs like this.
If you want to get out of line-of-business CRUD applications I think the key is specialization. You need to work one something niche and become a known contributor in that niche. It's more about who-you-know and being connected with the community in that niche.