LLM Research jobs

This is the center of gravity in AI hiring: pretraining and post-training research at companies where the model is the product. Pretraining seats are scarce and clustered at a handful of labs; post-training — RLHF and its successors, instruction tuning, reasoning — is where headcount has actually grown. “Member of Technical Staff” hides enormous variation, so ask what the team owns.

The daily work is experiments at scale: data ablations, reward model debugging, and reading loss curves the way traders read charts. JAX and PyTorch split the field, and everyone touches distributed training whether the title says engineer or scientist. Publication records help; shipped training runs help more; almost nobody prints the pay.