For companies
Serves large models fast and cheap, from kernels to clusters. We find the ones who can actually do it, and we figure out the right way to bring them to your problem.
An inference engineer makes models serve fast and cheap at scale. They work across the whole stack, from CUDA kernels up to cluster scheduling, and know which lever matters for the workload in front of them.
At scale every millisecond and every cent shows up in the bill. The job is cutting latency and cost without giving up quality, and knowing which of those the product actually needs.
Most hiring filters on credentials and years. The thing that makes a inference engineer good does not show up there. It shows up in how they work, which means you have to watch them work to see it.
That is what we do. We watch people work instead of reading resumes, so the person we send you is calibrated on the actual job, not the interview. Sometimes that is a hire. Sometimes it is a project or a person embedded for a while. We work out the shape with you.
They serve large models fast and cheaply, optimizing everything from GPU kernels to batching and cluster scheduling.
Comfort from CUDA up to Kubernetes, and the judgment to tell which optimizations are worth the complexity.
In the US, total compensation typically runs $220k to $340k, higher for kernel-level work at labs.