GPU Kernel Engineer — Fast ML Training & Inference

Tilde Research; Palo Alto (CA); On-site; USD 150,000 - 260,000; Full time

Job summary

Tilde Research is seeking a Kernel Engineer in Palo Alto to design, implement, and optimize high-performance GPU kernels for training and inference workloads. You will collaborate with ML researchers and engineers to push hardware limits, improve throughput, and reduce latency across models and infrastructure.

This role emphasizes performance-aware software, scalable kernels, and hands-on experimentation with CUDA, PyTorch, and Triton.

Qualifications

• Experience in deep learning or related research areas.
• Exceptional capability in ML kernel work (open source contributions, blogs, or similar).
• Familiarity with PyTorch, Triton/TK/TileLang (any 2+), CUDA, and GPU architecture.

Responsibilities

• Design, implement, and optimize GPU kernels for core model operations.
• Collaborate with ML researchers to prototype and scale model architectures.
• Contribute to system-wide efforts to improve efficiency and throughput beyond kernel-level work.

Skills

Deep learning

ML kernel development

PyTorch

CUDA familiarity

GPU architecture knowledge

Strong written communication

Quick learner

Problem solving

Tools

CUDA toolkit

Triton

TileLang

TK

Open source tooling

Job description

Tilde Research is seeking a Kernel Engineer in Palo Alto to design, implement, and optimize high-performance GPU kernels for training and inference workloads. You will collaborate with ML researchers and engineers to push hardware limits, improve throughput, and reduce latency across models and infrastructure.

This role emphasizes performance-aware software, scalable kernels, and hands-on experimentation with CUDA, PyTorch, and Triton.