Member of Technical Staff - Kernels & GPU Performance

Gimlet Labs; San Francisco (CA); On-site; USD 120,000 - 160,000; Full time

Job summary

A cutting-edge technology company in San Francisco is seeking a Member of Technical Staff focused on kernels and GPU performance. This role involves optimizing GPU and accelerator kernels for AI workloads by analyzing performance across various hardware. Ideal candidates have strong software engineering foundations and experience with performance-critical systems. Familiarity with tools like CUDA and performance profiling is preferred. This position offers a dynamic environment focused on real-world performance optimizations.

Qualifications

• Strong software engineering fundamentals.
• Experience working on performance-critical systems close to hardware.
• Comfort reasoning about low-level execution behavior and performance tradeoffs.

Responsibilities

• Design, implement, and optimize GPU and accelerator kernels for AI workloads.
• Analyze and tune performance across the GPU execution stack.
• Work with compilers and runtimes to ensure kernel performance.
• Bring up and optimize execution on new accelerators.
• Profile, benchmark, and debug performance issues.
• Ensure performance optimizations are production-ready.

Skills

Software engineering fundamentals

Performance-critical systems

Low-level execution behavior

Tools

CUDA

Triton

CUTLASS

Profiling tools

Job description

Gimlet Labs is building the first heterogeneous neocloud for AI workloads. As AI systems scale, the industry is hitting fundamental limits in power, capacity, and cost with today’s homogeneous, vertically integrated infrastructure. Gimlet addresses this by decoupling AI workloads from the underlying hardware. Our platform intelligently partitions workloads into components and orchestrates each component to hardware that best fits its performance and efficiency needs. This approach enables heterogeneous systems across multi-vendor and multi-generation hardware, including the latest emerging accelerators. These systems unlock step-function improvements in performance and cost efficiency at scale.

On top of this foundation, Gimlet is building a production-grade neocloud for agentic workloads. Customers use Gimlet to deploy and manage their workloads through stable, production-ready APIs, without having to reason about hardware selection, placement, or low-level performance optimization.

Gimlet works with foundation labs, hyperscalers, and AI native companies to power real production workloads built to scale to gigawatt-class AI datacenters.

Gimlet Labs is seeking a Member of Technical Staff focused on kernels and GPU performance. In this role, you will work close to accelerators and execution hardware to extract maximum performance from AI workloads across diverse and rapidly evolving platforms. You will analyze low-level execution behavior, design and optimize kernels, and ensure performance is reliable across both established and emerging hardware.

This role is ideal for engineers who enjoy deep performance work, reasoning about hardware tradeoffs, and turning theoretical peak performance into real-world results.

Responsibilities

• Design, implement, and optimize GPU and accelerator kernels for AI workloads

• Analyze and tune performance across the GPU execution stack, including memory access patterns, synchronization, and instruction scheduling

• Work with compilers and runtimes to ensure kernels integrate cleanly and perform well in end-to-end systems

• Bring up and optimize execution on new or emerging accelerators

• Profile, benchmark, and debug performance issues across kernels, runtimes, and hardware

• Ensure performance optimizations are robust, correct, and production-ready at scale

Qualifications

• Strong software engineering fundamentals

• Experience working on performance-critical systems close to hardware

• Comfort reasoning about low-level execution behavior, memory hierarchies, and performance tradeoffs

Preferred Qualifications

• Experience with CUDA, Triton, CUTLASS, or other accelerator programming models

• Deep understanding of GPU execution models (warps/wavefronts, blocks, grids)

• Experience optimizing memory access patterns (coalescing, shared memory, cache behavior)

• Familiarity with occupancy, latency hiding, and instruction-level parallelism

• Experience using profiling and performance analysis tools

• Familiarity with multi-GPU or distributed execution is a plus