Systems Software Engineer (C++)

Thunder Compute is a systems research lab with commercial focus, building GPU virtualization software. Our cloud revenue is growing 100%+ MoM and we are assembling a team of the world's top systems researchers and infrastructure engineers to change the way that GPUs are managed at scale.

Thunder Compute is the lowest cost, easiest GPU cloud for developers. We’re a 4-person team with $4.5M raised from top tier investors including Y Combinator and Matrix Partners.

There are two parts to our company. 1) we are a cloud platform where customers can rent GPUs just like they would from Coreweave, Lambda, AWS, etc., and 2) we are building systems software to use GPUs more efficiently than these competitors.

Your work will focus on core C++ systems development: performance optimization, systems debugging, and research.

You will spend your days bouncing between the weeds of complex, performance-critical systems in a production, customer-facing environment. This work is not easy - but if you’re the right candidate, this is a feature, not a bug.

• Top 0.1% C++ ability
• Experience optimizing NIC/GPU performance
• Deep understanding of low-level networking, compilers, and similar systems
• Experience in latency-sensitive environments

• You’ve got the dog in you: you will own your work and be responsible for production systems from day 1
• High speed-to-output
• Willingness to assume a senior leadership role as we scale
• Strong communication skills under pressure

• CS Bachelor’s degree from top 50 US university, 3.5+ GPA
• 1+ year of experience at a trading firm (Citadel, Jane Street, etc.), hardware company (NVIDIA, SambaNova), systems research or similar

• You'll report to Co-founder/CTO Brian (ex-Citadel quant dev)
• This role is full-time, in-person in SF

Our proprietary virtualization stack relies on our ability to network-attach GPU over TCP. We use a CUDA-compatible shim library to intercept calls to CUDA and optimize their execution across the data center. We look for deep expertise in systems architecture, low-latency optimization, and C++.

Our infrastructure uses Kubernetes as the foundation with custom multi-cloud networking, scheduling, and storage layers.

This is likely the hardest systems problem you'll find at any equivalently early stage startup.