Member of Technical Staff - Low Level & Kernels Capabilities

About the role

We are looking for experienced Machine Learning Engineers to join our Low Level / Kernels Capabilities team. The Kernels team focuses on developing reinforcement learning (RL) environments at the lowest layers of the stack, including GPU and accelerator kernels, vector ISAs, codec and crypto primitives, FPGA work, and other areas where frontier models are weakest.

Responsibilities

• Design and build low level / kernel-focused reinforcement learning (RL) environments that target a specified model and difficulty distribution.
• Choose which environments are worth building, ensuring they meet criteria such as targeting niche or genuinely hard domains, exercising real hardware features, and having a recognized reference to measure against.
• Develop robust, ungameable scoring mechanisms that are deterministic and can't be manipulated by models.
• Build correctness and performance scoring that is clear and unambiguous, requiring actual kernel development to achieve the objective.

Requirements

• Strong low-level/systems engineering skills: fluency in C/C++, CUDA (or an equivalent kernel language), and comfort with assembly when necessary.
• Proficiency in Python for engineering quality code, automation, and deployment scripts, data analysis, and plotting.
• Experience with hardware-aware coding, focusing on memory hierarchy, occupancy, data movement, parallelism, and trade-offs between latency and throughput.
• Kernel development experience, including iterative optimization against profiling data.
• An adversarial mindset, turning vague goals into robust, ungameable scoring mechanisms and anticipating how models might cheat.
• Ownership and autonomy, building, debugging, and shipping end-to-end systems with minimal supervision.
• Experience with large language models (LLMs) and ownership of their development and deployment.
• Depth in a niche hardware target or instruction set architecture (ISA), such as FPGA/HLS, RISC-V Vector, DSPs, SIMD/AVX, TPUs, or formal verification techniques like Lean or Coq.
• Open-source contributions and a strong competitive programming background, ideally in a low-level language.
• Experience with RL environments, agent harnesses, or evaluation infrastructure.

Qualifications

• Experience shipping a kernel that approached state-of-the-art (SOTA) and explaining the remaining gaps.
• Depth in a specific hardware target or ISA, such as FPGA/HLS, RISC-V Vector, DSPs, SIMD/AVX, TPUs, or formal verification techniques like Lean or Coq.
• Experience in adjacent disciplines such as HPC/heterogeneous clusters, hardware design (RTL/HDL, HLS), compilers and kernel toolchains (MLIR/LLVM, Mojo, Triton, gem5), or formal verification (Lean, Coq, SMT).
• The ability to read and apply performance and architecture papers to practical code.
• A strong competitive-programming background, ideally in a low-level language.

Benefits

• Competitive cash and equity compensation (up to 90th percentile)
• Ownership and autonomy in a fast-moving startup environment
• Opportunity to work with top machine learning engineers
• Health, vision, dental, and benefits
• 401K match
• Lunch provided every day onsite
• Weekly snack orders
• Visa sponsorship and relocation support available

Pay

$200K - $350K