Skip to Content
On-site

Performance Engineer, Kernels

Job Overview

We are hiring two specialized performance roles - Kernels (this posting) and Inference (companion posting). The Performance Engineering team owns the numbers the rest of the company plans against: how fast we serve, how much it costs, and how much we get out of every GPU.

Key points

  1. You will own the kernel layer.

  2. Where stock libraries - cuBLAS, cuDNN, FlashAttention, out-of-the-box Triton - leave performance on the table, you will author the custom CUDA, DSL-based…

  3. CUDA at kernel-authoring level: thread-block sizing, shared-memory layout, warp primitives, async copies (cp.async, TMA), and MMA selection.

  4. CUTLASS / CuTe DSL at a modify-and-extend level, with comfort in the layout algebra.

  5. PTX at a debug-and-modify level - you have inserted hand-written PTX where the compiler missed.

  6. Nsight Compute and Systems fluency: you read a roofline plot and propose the fix.

  7. Attention kernels: you have authored or modified at least one (FlashAttention-family, paged, MLA, sliding-window, or sparse).

  8. Multi-architecture awareness: what changes from Hopper to Blackwell (TMA, WGMMA, tcgen05).

Skills

Node.jsREST APIsGitMachine LearningLLMsCommunication

Experience: 5+ years · Job type: Full-time