One platform from the classroom to the analyst team.
Isolated workspaces, GPUs sized to the job, and controls that keep usage legitimate and costs predictable.
Architecture
Each region runs its own independent stack, so data and workloads stay in the region a customer selects.
Web & sign-in
Browser access to JupyterLab and VS Code. Sign-in via LMS (LTI 1.3), SAML or OIDC single sign-on.
Courses, teams & budgets
Templates, roles, quotas and hard spending caps. Every session is tied to a verified course or customer.
Isolated workspaces
Container-isolated sessions on autoscaling CPU and GPU node pools. Idle sessions are stopped automatically.
Region-pinned data
Encrypted persistent storage in the workspace region, with encrypted backups held by a second provider.
GPUs sized to the workload
Most coursework doesn't need a whole data-centre GPU. We match the hardware to the job.
| Workload | Typical hardware | Who uses it |
|---|---|---|
| Notebooks, data analysis, SQL | CPU instances | Every user, by default |
| Coursework: training small models, computer vision and NLP labs | Fractional NVIDIA L4 (shared between several students) | Students in GPU-enabled courses |
| Projects, theses, fine-tuning | Dedicated NVIDIA L4 or L40S | Advanced students, researchers, analyst teams |
| Peak weeks and deadlines | Additional cloud capacity on demand | Courses whose campus hardware is full |
GPU capacity is provisioned per region on Amazon Web Services, alongside optional on-campus workstations that KernelQuad manages.
Responsible use, built in
GPUs are valuable and attractive to abuse. These controls apply to every account.
Verified accounts
GPU access requires enrolment in a verified course or a paying team account. Anonymous sign-ups get CPU only.
Hard budget caps
Every user, course and team has a spending ceiling. Workloads pause when it's reached.
Workload monitoring
We detect and stop prohibited workloads such as cryptocurrency mining, scanning and spam, as set out in our acceptable use policy.
Session limits
Idle sessions shut down, long-running jobs need an explicit schedule, and network egress is restricted by default.
Tools you already know
Languages
Python, R, SQL and Julia, with conda and pip package management.
Frameworks
PyTorch, TensorFlow, JAX, scikit-learn, pandas, Polars and Spark.
Interfaces
JupyterLab, VS Code in the browser and a terminal — all in one workspace.
Data
Upload files, mount shared course datasets, or connect databases and warehouses.
See it with your course or team
We'll set up a pilot workspace in the region you need.