I'm a team lead engineer at Annapurna Labs, the small company-within-a-company that designs AWS custom silicon. My organization designs and runs the Trainium and Inferentia machine learning (ML) accelerators. Our team is hiring engineers to help build the on-chip firmware stack for new and existing chips. This firmware is the lowest layer of the Neuron software stack; you can think of it as implementing a silicon-accelerated virtual machine that the Neuron compiler targets.
This is a bare metal, system software-oriented role (think C/C++); no specific ML experience is needed, but if you have experience writing GPU/CUDA kernels or microcode for other systems, we can definitely put your experience to work. If you're interested in machine learning, you'll be adjacent to people who have tons of experience with high performance compute, chip design, and ML algorithms.
We offer relocation to Austin or Cupertino; the team is split evenly between the two sites.
Ping me if you're wanting to chat about the team or the role: ghilliar -at- amazon.com. I'm an embedded systems engineer and I don't bite :)
And here's the job description link: https://www.amazon.jobs/jobs/2894488