Hypervisor role vs ML compiler startup
I am a fresh grad and deciding between two paths.
One is a systems role at Siemens DISW involving hypervisor virtualization Intel Arc and OS level work.
The other is a startup doing ML optimization or compiler related work on three layers
I love low level topics and never really got in ML stuff. I like compilers OS internals architecture performance etc. I have read compiler material and built a small AOT compiler before.
My concern with Siemens is ending up in slow corporate maintenance work especially it being in a third world regional office.
My concern with the startup is that the compiler side may mostly be AI hype or infra glue instead of real systems or optimization work. They do claim to be optimizing across all three layers Abstract layer (Quantization etc), MLIR (Compiler) and SIMD compute kernels. Engineers over there have max 4 YOE not just in ML compilers/optimizations but like as in total experience. Although they did start with this work.
Long term I care more about durable technical depth and hard to automate skills than hype.
For people working in systems compilers virtualization or AI infra
Which path do you think ages better over 5 to 10 years
How transferable is hypervisor and systems experience into other fields
How do you identify whether an ML compiler startup is doing real technical work or mostly marketing and if they will bait and switch me into high level ML stuff (a big no no)
How saturated or expected to be saturated is ML compiler Path
I am trying to optimize more for long term technical growth than immediate salary.
Thanks in advance for your advice