FlashAttention-3 Jobs

2 FlashAttention-3 jobs currently listed

Browse 2 open positions that require FlashAttention-3 experience. Every listing includes the full tech stack, language requirements, remote policy, and salary data, so you can filter with confidence instead of guessing from job descriptions.

By the numbers

  • 2 companies hiring
  • Mostly Senior roles
  • Often paired with Python, PyTorch, CUDA, NCCL
  • Top locations: Amsterdam

What you'll find here

Each job includes structured data from the original posting. You can see exactly which technologies are required vs. nice-to-have, whether the role is remote-friendly, and what salary range to expect, all without reading through pages of job descriptions.

JetBrains 2 months ago
On-site

Research Engineer (LLM Training and Performance)

Amsterdam, Netherlands +10 - English

Python C++ PyTorch PyTorch Distributed Megatron-LM +12
Firstprinciples 3 months ago Open for 118+ days
Remote

Member of Technical Staff, Training Engineer (Large Scale Foundation Models)

English

PyTorch Python CUDA Triton FSDP +7

Frequently Asked Questions

How many FlashAttention-3 jobs are available right now?
There are currently 2 open positions that list FlashAttention-3 as a required skill. This number updates daily as new jobs are posted.
How do you know a job requires FlashAttention-3?
We analyze every job description to identify the specific technologies mentioned. We distinguish between technologies that are required, nice-to-have, or explicitly not used, so you get an accurate picture of each role's tech stack.
Can I filter out jobs that use certain technologies?
Yes. Jobdex supports negative filtering, which lets you exclude specific technologies from your search. For example, you can find FlashAttention-3 jobs that don't use a particular framework or language.
Are salary ranges accurate?
When a company includes salary in their job posting, we use that directly. Otherwise, we parse salary information from the job description. Each salary figure gets a confidence score. High-confidence data is shown prominently, while less certain figures are clearly marked.