Back to Jobs
Full timeOther

Member of Technical Staff

WaferSan Francisco🇺🇸United States

Why this role stands out

This role offers incredible growth potential by allowing you to directly impact customer success and own critical systems from development to production. You'll thrive here if you're eager to tackle complex challenges in AI inference optimization and build direct relationships with innovative AI startups. Apply now to be at the forefront of cutting-edge hardware optimization!

Quick overview

Salary$200k - $300k/yr
SeniorityMid Senior
Employment typeFull Time
Work modeOn Site
LocationSan Francisco, United States
CUDALLM

Job description

Member of Technical Staff

At Wafer, we're building AI systems that automatically optimize inference workloads across silicon. The goal is fungible token capacity: any accelerator, optimized to serve inference as efficiently as possible.

Inference performance depends on how well models use the hardware underneath them. Gains across kernels, compilers, runtimes and serving systems can be dramatic, but they're hard-won, hardware-specific and slow to reproduce on every new accelerator. We're automating that work, and running it in production. Wafer serves trillions of tokens for mission-critical workloads at fast-growing AI startups, so what you build is live in real traffic, not a benchmark.

Members of Technical Staff build the core of Wafer: the kernels, inference engine, optimization agents and infrastructure behind every workload we serve. Our forward deployed engineers run customer deployments and bring you the hardest problems they find.

What you'll do

  • Write and optimize high-performance kernels, and work across inference engine internals and serving infrastructure.

  • Build AI agents that do inference engineering autonomously.

  • Design, deploy and operate heterogeneous clusters across hardware vendors.

  • Solve the hardest problems that come out of production, when a workload needs a fix in the engine, the kernels or the agents.

  • Shape the roadmap from what you see in the system under real load.

What we look for

  • You're an exceptional engineer who has shipped and operated production systems, and can still open a profiler, read a trace and find the problem yourself.

  • Depth in at least one of: GPU kernels (CUDA, Triton, ROCm or similar), inference engines (vLLM, SGLang, TensorRT-LLM), ML compilers, or large-scale distributed systems. Inference experience helps, but it matters more that you can take an unfamiliar system and own it within a week.

  • You work without a spec. You take a half-defined problem and come back with a scoped answer, not a list of questions.

Why Wafer

  • The problems are at the frontier: making models run fast on whatever silicon exists, automatically.

  • Your work ships to production traffic at scale, quickly.

  • No narrow lanes. You'll work across the stack at a company scaling extremely fast.

How we evaluate

We score every candidate on seven values:

  1. Infinitely Resourceful

  2. Exceptionalism

  3. Unreasonable Standards

  4. Company Over Self

  5. High EQ

  6. Learns Quickly

  7. First Principles Thinker

Compensation and benefits

  • $250K + generous equity.

  • Fully covered medical, dental, and vision insurance.

  • Daily lunch and dinner, unlimited PTO, and parental leave.

  • $1K/month housing stipend (post-tax) if you live within walking distance (0.5 miles) from the office.

  • Covered Uber/Waymo from/to office.

  • Relocation allowance and benefits.

  • Visa sponsorship available.

How we work

On-site in San Francisco, five days a week. Small team with massive surface area and ownership. We don't see engineers as code writers but as problem solvers. Nearly all of your time is engineering, not meetings.

Haystack's AI agent can fill in and submit this application for you. How AI auto apply works

Similar jobs