Curtin Institute for Data Science

CIDS facilities and resources

/facilities/PerthAU

Facilities & Resources

Compute, tooling and secure environments for data-intensive work.

CIDS helps teams choose and use the right infrastructure for the work: local AI training GPUs, high-memory compute, national HPC pathways, cloud patterns, research software support and secure spaces for sensitive data.

Local

Curtin AI GPU access and Perth-based workshops

National

Pawsey support, allocation planning and HPC advice

Hybrid

Cloud, secure data and reproducible project handover

1. Infrastructure

What infrastructure do you have?

A practical mix of people, platforms and pathways for research, industry and government projects that need more than a laptop.

HPC pathways

Project-level support for parallel workloads, batch scheduling, node-scale runs and national supercomputing access.

GPU training access

Facilities access to Curtin on-prem AI training GPUs, with advice on when to use local, cloud or Pawsey resources.

High-memory compute

Options for workloads that need larger RAM footprints, long-running jobs or data-heavy preprocessing before modelling.

Research software support

Help to refactor slow scripts, package methods, containerise workflows and make analysis reproducible for collaborators.

Secure project spaces

Controlled environments for sensitive datasets, approvals-aware collaboration and clean handover of code and outputs.

Visualisation rooms

Perth-based collaboration spaces for project discovery, dashboard review, technical workshops and partner demonstrations.

Evidence Gallery

Data-first project previews for outcome-led case studies and dashboards.

Data
A research workstation prepared for a production computing workflow.

Parallelise a thing

Turn serial research code into scheduled jobs, multi-process pipelines or node-aware workflows that can use more of the machine.

A camera and computing workstation used for machine learning prototyping.

GPUify a thing

Move eligible ML, simulation or matrix-heavy workloads onto CUDA-enabled tools, GPU libraries and training environments.

Large analytical displays used to monitor computational outcomes.

Across nodes

Plan data movement, job arrays, MPI-style execution and result collection across nodes in a supercomputer environment.

Researchers preparing a technical pathway together at a whiteboard.

Pawsey fast track

Use CIDS as a preliminary step for Pawsey fast-track conversations, allocation planning and general supercomputing support.

Monitoring equipment connected to an industrial data workflow.

GPU and high-mem availability

Match the workload to Curtin on-prem AI training GPUs, high-memory machines, national infrastructure or cloud capacity.

Data governance notes and schema material spread across a project desk.

Cloud when it fits

Design cloud runs for burst capacity, managed services, deployment experiments or partner-controlled environments.

3. AI/ML platforms & tooling

A stack for experiments that can grow up.

The tooling mix depends on the project, but the aim is consistent: reproducible environments, accelerated modelling, inspectable data pipelines and handover-ready software.

6 platform groups

Model development

Notebook-to-package workflows for experimental and applied ML.

AI/ML
PythonRJuliaPyTorchTensorFlowscikit-learn

Accelerated computing

GPU-aware libraries and profiling for heavier numerical work.

GPU
CUDARAPIDSNumbaCuPyJAXNVIDIA containers

Workflow orchestration

Repeatable pipelines that can move from laptop to cluster.

Pipelines
SlurmNextflowSnakemakeDVCAirflowGitHub Actions

Research environments

Interactive workspaces for mixed technical teams.

Workspace
JupyterHubRStudioVS Code ServerCondauvPoetry

Data platforms

Storage, query and feature preparation for analysis-ready data.

Data
PostgreSQLDuckDBSparkParquetPostGISS3 patterns

MLOps and delivery

Tracking, serving and monitoring for models that need to keep working.

Delivery
MLflowWeights & BiasesFastAPIDockerApptainerPrometheus

4. Secure / sensitive data environments

Controls that fit the data, approval and partner risk.

Multiple controls, one project pathway

Access control

Controlled

Project workspace

Role-based permissions, Curtin identity patterns, MFA-ready access and least-privilege collaboration.

Ethics-aware data handling

Governed

Research governance

Project setup can reflect ethics approvals, consent limits, retention requirements and participant-risk controls.

De-identification workflows

Private

Sensitive data

Support for removing, masking or separating identifiers before analysis, modelling or external collaboration.

Encrypted storage

Protected

Infrastructure

Patterns for protected storage, transfer, backup and controlled project handover of source data and outputs.

Audit trails

Traceable

Delivery controls

Versioned code, documented data movement, reproducible runs and reviewable decisions for sensitive projects.

Secure cloud patterns

Hybrid

Hybrid delivery

Advice on when approved cloud services, private networking, key management and local compute are the right fit.

5. Can we use it for our project?

Usually, yes, but the right setup depends on the work.

Some projects need a short burst of GPU capacity, some need a secure analysis environment, some need a Pawsey allocation, and some just need someone to look at a slow script and make the next step obvious. CIDS can help work out which path is realistic before the project becomes bigger than it needs to be.

Bring the research question, dataset, model idea or operational problem and we can help assess compute fit, security constraints, software effort, budget shape and whether the work should run on Curtin infrastructure, Pawsey, cloud services or a blended setup.

Step 1

Scope

Clarify data, risk, compute and expected outputs.

Step 2

Match

Choose the environment and support model.

Step 3

Deliver

Build, run, document and hand over the workflow.