Familiar by design
The batch workflow your people already know. Existing job scripts run as-is; zero retraining.
Your own HPC cluster. Without the cluster.
Podulus gives labs, courses and workshops a private high-performance computing environment — login node, batch jobs, GPUs on demand — ready in a day, billed only for what you actually compute.
Shared clusters queue longest exactly when deadlines hit.
Students and external attendees wait weeks for accounts.
Dedicated hardware idles — and still needs someone to run it.
features
Everything a working cluster needs. Nothing to run yourself.
Jobs start in seconds
Capacity follows demand. Forty submissions at once all start now — deadline night is just another night.
GPU & CPU partitions
Debug tier through production-class GPUs, plus large-memory CPU nodes for solvers and pre/post-processing.
You're in control
Add members in one click, set per-member hour caps, watch a live usage burn-down.
Hours that never expire
Prepaid GPU-hour packs carry over indefinitely. No use-it-or-lose-it cycles, no surprise invoices.
Software included
Curated module tree with common compilers and solvers. Requests handled within a working day.
Deadline night is just another night.
On a shared cluster, everyone submits at once and everyone waits. Here, capacity follows the queue: the fortieth job gets a node just as fast as the first, and nothing sits idle between deadlines.
[maria@login ~]$ squeue --me
| JOBID | PARTITION | NAME | ST | TIME | NODES |
|---|---|---|---|---|---|
| 1042 | gpu | run_case | R | 0:12 | 1 |
| 1043 | gpu | mesh_prep | R | 0:11 | 1 |
| 1044 | gpu-debug | sweep_01 | R | 0:11 | 1 |
| 1045 | cpu-lm | solver_a | R | 0:10 | 2 |
| 1046 | gpu | train_res | R | 0:09 | 1 |
| 1047 | gpu | sweep_02 | R | 0:08 | 1 |
| 1048 | cpu-lm | postproc | R | 0:07 | 1 |
| 1049 | gpu | train_vit | R | 0:06 | 2 |
| 1050 | gpu-debug | notebook | R | 0:05 | 1 |
| 1051 | gpu | sweep_03 | R | 0:04 | 1 |
| 1052 | cpu-lm | assemble | R | 0:03 | 4 |
| 1053 | gpu | run_case2 | R | 0:02 | 1 |
how-it-works
From nothing to running jobs in a day.
-
01
We set up your cluster pod
Login node, storage, software, member accounts. Live in a day.
-
02
Your people ssh in and submit
Batch jobs, interactive GPU sessions, personal scratch storage with direct high-speed transfer.
-
03
You watch the dashboard
Usage, costs and caps in real time; top up hours whenever you need them.
who-its-for
Built for the way research and teaching actually run.
Research labs
A private cluster for up to 8 people. Bursty production runs without renting idle hardware.
from $300/month + compute packs
Courses
Real cluster experience and transferable batch-computing skills. Per-student caps and instructor reports.
flat per-course pricing
Workshops & summer schools
External-attendee accounts in minutes, your code pre-installed, burst capacity in hands-on sessions. Archived and gone afterwards.
flat per-event pricing
EU-resident storage available. Your data stays yours — direct download credentials, standard formats, no lock-in.
pilot
Pilot with us — compute on us.
We're onboarding a small number of pilot labs, courses and workshops. You bring the users; we fund the pilot's compute hours. If it doesn't remove real headache, walk away owing nothing.
Prefer to talk first? hello@podulus.com
faq
Questions we get asked.
Do we have to rewrite our job scripts?
No. You get the same batch workflow your group already uses — submit with a job script, check the
queue, collect results. Existing scripts run as-is, and anyone who has used a university cluster is
productive on day one.
How fast can we actually start?
A working day. We set up the login node, storage, software modules and member accounts, then send
you the connection details. Adding people afterwards takes a click each — including external
workshop attendees who have no institutional account.
What does it cost, and what happens to unused hours?
A small monthly base for the environment, from $300 for a lab, plus prepaid compute packs you draw
against. Hours never expire and carry over indefinitely — no use-it-or-lose-it cycles. Courses and
workshops are priced flat, per course or per event.
Where does our data live, and can we get it out?
EU-resident storage is available on request. Your data stays yours: direct download credentials,
standard formats, no proprietary containers and no lock-in. Tell us to delete an environment and it
goes, along with everything in it.
We already have a cluster. Why would we want this?
Most groups keep theirs. Podulus covers the parts a shared cluster handles badly: the deadline
crush when queues are longest, a course that needs thirty accounts for one semester, a workshop
that needs GPUs for two days, or a run that needs a GPU your cluster doesn't have.