Please mention DailyRemote when applying
See how much of this job your resume covers, and what’s missing.
Want a recruiter to go through it line by line?
Get professional reviewOne click makes a copy of your resume for this role, scores it against the posting, and proposes edits line by line. Your default resume is never changed.
Upload your resume and we draft a letter for this exact role, tailored to what it asks for.
You will build and manage Kubernetes and GPU virtual machine products on bare metal infrastructure. You are responsible for the entire lifecycle of these services, including service levels, monitoring, and customer-facing operations.
The limit on superintelligence (SI) is no longer chips. It's power. New data centers wait years for a grid connection, while the office towers, hotels and hospitals of every major city sit on power they were built for and no longer use. Perimeter Compute is on a mission to put that power to work.
We turn basements and mechanical rooms into GPU clusters, and our software links them into one inference network running inside the cities where SI will be used. Building owners earn from space that would otherwise sit idle. Frontier labs and inference providers get capacity without the years-long wait, a few miles from their users instead of a few states away.
Perimeter is a small, highly leveraged team that has worked every layer of this problem: building GPU infrastructure at Lambda and HPE, engineering commercial buildings at Tishman Speyer and Trane, and routing inference at Subquadratic. The power is already there. Come build what runs on it.
People make the decisions that matter. Agents do everything else.
Agents write and review all of our code.
People own the calls that are hard to undo: architecture, product direction, taste, and the frameworks we use to make those calls.
Most decisions are two-way doors, so you make the call instead of waiting for one.
If you like writing code yourself, you'll hate it here.
One owner per outcome. You build it, you run it, you grow it.
Roles here are bigger than usual. You'll do work that's normally split across engineering, product, sales and design.
You'll work directly with customers to figure out what to build.
You own the outcome, so you decide how to get there.
If you want a clear ticket and someone else owning the result, this isn't the job.
We keep reinventing how software gets built.
Code is a means to an end, not the end.
We question every assumption. "Why?" is the most important question you can ask, and a rule nobody can answer it for gets deleted.
How we build is code too. Every week, agents find what slowed us down and ship the fix.
If you want a process that stays the same, look elsewhere.
We build for agents as much as for people.
Agents use our tools, APIs, docs and interfaces too. It isn't done until an agent can use it and recover from its errors without help.
Every public API is defined once in code, and everything people and agents read about it is generated from that, so nothing drifts.
Agents don't remember rules, so anything that matters is a check that fails, not a line in a doc.
Because agents do the work, we can be prescriptive: one way to build things, and every agent follows it.
If docs and error messages feel like someone else's job, we're not a fit.
You'll own the layer right above our bare metal: the products customers will rent. Our GPUs are spread across buildings, but customers should rent one cloud: a node failing or a site losing its link is our problem, never theirs. You'll build managed Kubernetes and GPU virtual machines on our own hardware, starting with Kubernetes, then own them as products: their service levels, monitoring, adoption and direction. We're building to the criteria behind the top tier of ClusterMAX, SemiAnalysis's public rating of GPU clouds.
You're measured on paying customers and how much of our fleet is ready to rent at the service levels we publish.
Stand up managed Kubernetes on our bare metal from scratch, with hosted control planes and the GPU and networking stack customers expect
Build GPU virtual machines: whole nodes first, then smaller sizes where customers need them
Turn one pool of capacity into reserved, on-demand and spot products, with the quota and preemption rules behind them and the usage records billing is built from
Build isolation between customers into everything you ship, to the bar the security team sets and tests
Make failures ours, not the customer's: move work off a failing node, and keep workloads running when a site loses its link to our central control plane
Make every operation work for agents as well as people, through our API, CLI and agent tools
Run what you ship: publish its service levels, get paged when they slip, and roll out Kubernetes and GPU driver upgrades without breaking customer workloads
Talk to the developers using it and decide what to build next
Have built and run production software that people depend on, and owned it when it broke, whether that took you two years or ten
Have run production Kubernetes on bare metal, including provisioning, upgrades and node operations
Have debugged GPU failures in production: drivers, Xid errors, NCCL hangs
Have working knowledge of KVM/QEMU, KubeVirt and GPU passthrough
Have owned a product other developers paid for or depended on, including its service levels
Get real results from coding agents on infrastructure, and know which changes a person must check
Hosted control planes with Gardener or k0smotron
Bare-metal provisioning with Ironic, Metal3 and Redfish
Network fabrics like Netris or Hedgehog, and BlueField DPUs
GPU partitioning: MIG, or NVLink partitions through Fabric Manager
Quotas and preemption with Kueue
Fully remote in the US. We aren't able to sponsor visas at this time.
Health, dental and vision insurance
Unlimited PTO
A $500 home office stipend
A top-spec MacBook
Unlimited tokens
Don't meet every bullet? Apply anyway. We care more about what you've built and how you think than about checking every box.
Please write your answers to the application questions yourself, without AI. After you apply, we'll tell you how to use AI in each interview round.
We accept applications for this role on an ongoing basis.
Perimeter Compute is an equal opportunity employer. We don't discriminate based on race, color, religion, sex, pregnancy, sexual orientation, gender identity, national origin, age, disability, genetic information, veteran status or any other legally protected status. If you need an accommodation at any point in our hiring process, email careers@perimetercompute.com.
Stop the endless job search. Our AI finds and applies to the best jobs for you.
Featuring 222,035+ Jobs in Others
Answer easy questions
222,035+ jobs across 15+ categories
Get your best job matches
Only hand-screened, legit jobs
Find a remote job faster
No ads, scams, or junk
“I was the first applicant for a remote marketing position that got listed on the company website the same day I applied. Had an interview within 48 hours!”