Application

Everything you write and record must be in English. The video is optional. Fields marked * are required.

Full name *

Email *

LinkedIn profile *

Where do you live? *

This role is based in San Francisco or Seattle.

U.S. work authorization *

If you hold a visa that will need sponsorship later, such as OPT or H-1B, choose the second option.

Demo video link (optional)

Five minutes or less, in English. Loom, YouTube, Vimeo, or Google Drive (unlisted is fine), and the link must point at the video itself, not a folder. Make sure it is viewable by anyone. If you record one: your face on camera introducing yourself, and a screen recording walking through a serving system you built or tuned and the numbers it hit. Casual is fine, content matters far more than polish.

No video? Leave the field blank and let your written answers carry the application.

Rate your English, 1 to 6 *

We are looking for 4 or higher. Answer honestly, the interview will show it anyway.

Rate yourself 1 to 10 *

vLLM and SGLang

NVIDIA Dynamo and distributed serving (disaggregated prefill/decode, KV-aware routing)

GPU performance: profiling and kernels (CUDA, ROCm, Triton)

Quantization and model formats (FP8, FP4)

Python and PyTorch

What is the hardest serving problem you have solved in production, and what did it move? *

Numbers you measured yourself. This is the answer we weigh most heavily.

Have you contributed to vLLM, SGLang, Dynamo, or another inference engine? *

How do you use AI agents in your engineering work? *