Member of Technical Staff

Company: Confidential - Early-stage AI cloud startup
Location: San Francisco, CA (FiDi office, on-site 5 days per week)
Compensation: $200,000 - $300,000 + 0.5% - 1% equity
Employment Type: Full-time
Visa Sponsorship: Visa transfers for candidates already in the US (no new H-1B or overseas sponsorship)

About the Company

An early-stage AI cloud startup maximizing intelligence per watt, serving open-source LLMs using autonomous agents that write and tune GPU kernels across heterogeneous hardware.

The Role

Our client is hiring a Member of Technical Staff (1-6 years) to work close to the hardware on its inference stack. On a small team with massive surface area, you will do everything from talking to customers (10-20% of the role) to writing custom GPU kernels for esoteric hardware, with full autonomy over how you solve problems and a hand in setting the direction of its inference serving infrastructure.

What You Will Do

• Ship day-zero support for new open-source models, tuned for latency and throughput.
• Optimize the serving stack: batching, KV cache, speculative decoding and quantization.
• Write and tune kernels in CUDA, HIP and Triton for NVIDIA, AMD, TPU, Trainium and other accelerators.
• Design, deploy and operate heterogeneous clusters across vendors.
• Run production inference across a mixed fleet with strong reliability, observability and cost per token.

What You Bring

• 1-6 years of software engineering focused on systems close to the hardware
• Deep understanding of computer architecture, memory hierarchy and OS internals
• Backend, infrastructure or systems work at a top-tier tech company, hardware-adjacent company, quant firm or strong AI startup
• CS degree from a top undergraduate program
• Evidence of exceptionalism (competitions, rankings, standout projects)
• Enthusiasm for AI agents and coding tools in your own workflow
• High EQ and clear communication with customers
• Ability to work on-site in San Francisco 5 days a week

Nice to Have

• New grads with strong internships are considered
• Deploying or operating infrastructure at scale (data centers, clusters, GPU fleets)
• ML inference optimization (quantization, batching, KV cache, speculative decoding)
• CUDA, Triton or GPU programming; OS, architecture or compilers coursework

Benefits

$1,000/month housing stipend for anyone living within half a mile of the office.

Interview Process

Intro screen (20 min), combined founder screen and technical round, co-founder screen (20 min), technical interview (45 min), one-day on-site work trial.

Tech Stack

CUDA, HIP, Triton, Python, TPU, AWS Trainium