Skip to content
State of GPU Compute 2026

Careers

Build the cloud you'd rent yourself.

A small team running whole 8-GPU servers for AI teams across Europe and the US. Hardware, software and the invoice, all in-house.
View open rolesHow we work
nexpod
LD

Team member

Platform

RegionEU-AMS-1Badge#0142
nexpod
RN

Team member

Data center ops

RegionEU-AMS-1Badge#0087
nexpod
EC

Team member

Solutions

RegionUS-NYC-1Badge#0219
Placeholder
open roles
16open rolessample list
teams hiring
4teams hiringEng · Infra · GTM · Ops
office hubs
3office hubsAMS · FRA · NYC
core overlap
2 hcore overlap15:00–17:00 CET

Why it matters

GPUs are the bottleneck. We remove it.

AI teams lose weeks waiting for capacity, fighting quotas and decoding bills. Every server we rack and every second we meter correctly gives some of that time back.

Our mission

Make a whole GPU server as easy to rent as a VM, and as honest to pay for.

How we work

A normal week. Written down.

Four habits shape most weeks here, from the rack floor to the review doc. None of them need a meeting.
EU-AMS-1 · Hall B · Rack 07 · U12 sled out
GIGABYTE G894-ZD3-AAX7
GPU 5 · reseat and burn-in · 40 minPhoto placeholder

Hardware you can touch

We run our own racks. Engineers visit them, pull a sled now and then, and design for the real failure modes.

docs/scheduler/placement-v3.mdapproved

Placement v3: boot-aware scoring

## Failure modes

TAWhat if a node fails mid-boot?
LDAdded a retry path under ## Failure modes

Own a system, end to end

The design doc, the rollout, the dashboard and the page when it breaks. One owner, many reviewers.

On-call · Q41 week in 6
W40SB
W41YOU
W42TA
W43LD
W44RN
W45MK
W46SB
W47YOU
W48TA
W49LD
W50RN
W51MK
Night page in W41→ paid day off

On-call that respects sleep

One week in six, and a paid day off after a rough night. Every page gets a follow-up.

One day · CET core 15:00–17:00

Amsterdam

CET

New York

ET

0812162024

Meetings

2 / week

Updates

written

Overlap

2 h

Async by default

Two shared core hours, written updates for the rest. Meetings need an agenda, and most end early.

Benefits

Sample terms

What you get. In writing.

The same package for every hire at a level, in every country we employ in. No negotiation games.
Engineering · bandsshared before interview 1
L1
L2
L3
L4
L5
minmidmax

Published salary bands

You see the band for your level before the first interview, not after the offer.

Options · vesting4 yrs

Equity for everyone

Options for every permanent hire. Four years, one-year cliff, monthly after.

August7 of 30 used
MTWTFSS01020304050607080910111213141516171819202122232425262728293031

30 days off

Plus local public holidays. Two weeks in a row, every year, is expected.

Credits / month€300
B300
50 GPU·h
B200
60 GPU·h
H200
75 GPU·h
H100
100 GPU·h

Your own GPU budget

Monthly credits on the production fleet for side projects and experiments.

  • Laptop of your choiceday 1
  • Monitor, chair, deskshipped
  • Co-working passif remote
  • Pension + health coverlocal

Set up properly

The laptop you want, a real chair and a desk that fits your room.

Amsterdam

Data center day + planning

April

5 days

Somewhere warm

Hack week, no roadmaps

October

5 days

Together twice a year

Two offsites, travel paid. One for planning, one for building.

Learning · this year€1,220 of €1,500
Conference · ticket + train€640
Course · CUDA programming€420
Books × 4€160

Learning budget

A yearly budget for courses, books and conferences. No approval form under €500.

Open roles

Sample roles

Come build it with us.

Ownership from day one, real hardware, and a front seat as GPU clouds grow up.

Engineering 7

7 roles
  • ML Systems Engineer, Serverless InferenceNew
    serverlessEuropeRemote (EU)

    Cut cold starts and queue time for Serverless endpoints. Profilers, schedulers and a lot of graphs.

    Apply
  • Senior Platform Engineer, SchedulerNew
    platformAmsterdamHybrid

    Own the placement engine that decides which 8-GPU node a request lands on, and how fast it boots.

    Apply
  • Site Reliability Engineer
    reliabilityFrankfurtHybrid

    Keep three regions boring. On-call one week in six, with time after each incident to fix the cause.

    Apply
  • Developer Experience Engineer, nexctl
    devexEuropeRemote (EU)

    Own nexctl, the SDKs and the docs examples. Every command should work the first time.

    Apply
  • Systems Engineer, GPU Drivers & Firmware
    hostEuropeRemote (EU)

    Driver, firmware and kernel work across B300, B200, H200 and H100 hosts, from burn-in to rollout.

    Apply
  • Staff Engineer, Networking & Fabric
    fabricAmsterdamHybrid

    Design the 400G fabric behind Instant Clusters and make multi-node jobs see one flat network.

    Apply
  • Product Engineer, Console
    consoleAmsterdamHybrid

    Build the console customers live in: server lists, usage, billing. TypeScript, React, taste.

    Apply

Infrastructure 4

4 roles
  • Data Center TechnicianNew
    dc-opsAmsterdamOn-site

    Rack, cable, burn in and repair GPU servers in EU-AMS-1. Shifts, tooling and spares included.

    Apply
  • Capacity Planner
    fleetAmsterdamHybrid

    Forecast demand per GPU family and region, and decide what we rack next quarter.

    Apply
  • Data Center Technician
    dc-opsNew YorkOn-site

    The same hands-on work in US-NYC-1: installs, swaps and the occasional 3 a.m. page.

    Apply
  • Hardware Lifecycle Lead
    fleetFrankfurtHybrid

    From purchase order to retirement: vendors, burn-in plans, spares and RMA flow for the whole fleet.

    Apply

Go-to-market 3

3 roles
  • Solutions Engineer, EMEA
    solutionsEuropeRemote (EU)

    Size clusters with customers, run proofs of concept and turn workloads into quotes in euros.

    Apply
  • Account Executive, DACH
    salesFrankfurtHybrid

    Reserved and cluster deals with teams across Germany, Austria and Switzerland. German required.

    Apply
  • Account Executive, US East
    salesNew YorkRemote (US)

    Grow US-NYC-1 with AI teams on the east coast. Technical buyers, short sales cycles.

    Apply

Operations 2

2 roles
  • Billing Operations Analyst
    financeAmsterdamHybrid

    Per-second usage in, clean euro invoices out. VAT, reverse charge and the edge cases between.

    Apply
  • Security Engineer, Infrastructure
    securityEuropeRemote (EU)

    Tenant isolation on bare metal, secure boot, audits. Make 'one tenant per node' provable.

    Apply

Hiring

Four steps, about three weeks.

Feedback after every step, even when the answer is no.
  1. 01
    30 min

    Intro call

    Thirty minutes on what you want next and what the role is.

  2. 02
    60 min

    Technical conversation

    A real problem from our backlog, talked through together.

  3. 03
    ≤ 4 h

    Paid work sample

    A scoped task at home, paid at your future day rate.

  4. 04
    Half day

    Meet the team

    Lunch or a call with the people you would work with.

Light reading

How we think, in writing.

Posts from the people you would work with: billing internals, data residency and hardware trade-offs.
All posts
00:47:12billed 2,832 s · B300€0.01333 / s€37.76
Billing6 min read
Billing team

Per-second billing, explained

When the meter starts, when it stops, and what a 47-minute job on a B300 server actually costs.

EU BOUNDARYEU-AMS-1EU-FRA-1EU-HEL-1OUTSIDEregistry
Security7 min read
Security team

Keeping training data inside the EU

How region pinning works, what stays in the region you choose, and one setting to check before you upload anything.

180GB · B200288GB · B300B200B300
Hardware9 min read
Infrastructure team

Choosing between B200 and B300 for 70B inference

Both fit a 70B model on one GPU. What differs is the KV cache left over, and that sets how many users each GPU can serve.

FAQ

Asked a lot.

Anything else? The recruiting team reads every message.
Where is the team based?
Most of us live near Amsterdam, Frankfurt or New York, close to our data centers. Engineering roles marked Remote (EU) can be done from any EU country with a CET or EET time zone.
Do you sponsor visas?
For roles in the Netherlands, usually yes, through the highly skilled migrant route. For other countries we decide per role.
What does the interview process look like?
Four steps over about three weeks: an intro call, a technical conversation, a paid work sample and a meeting with the team. You get feedback after every step.
Can I apply for more than one role?
Yes. Pick the one that fits best and mention the others in your note. We would rather move you between pipelines than make you start again.

Open application

Nothing open that fits? Write anyway.

Tell us what you would build here in a few lines. We read every note and keep it on file for a year.

Send an open applicationAbout Nexpod
open-application.md3 lines is enough
  1. 01What you would buildtwo or three lines
  2. 02Something you shippeda link is enough
  3. 03Where you want to workAMS · FRA · NYC · remote EU
Kept on file for 12 monthsread by a human