Research-to-production bridge

Taking a research group's working prototype, built to answer one question in one dataset, and hardening it into a tool their wider team could run without a developer present.

Timeline: August 2024 - February 2025
Runtime: 6 hours to 20 minutes
Team members now able to run it: 5

A research desk and a production workstation shown side by side. — Image adapted from the CIDS project image set for the project detail page.

It went from something only one of us could run to something the whole team uses.

RG

Research group lead

Curtin research partner

Recognition and proof

Project signals

Researchers working through diagrams during a funded discovery project.

Performance

Cut processing time by 95%

The original prototype's single-threaded analysis pipeline was rebuilt to process data in parallel, taking a six-hour run down to twenty minutes.

An overhead camera rig inspecting parts on a workbench.

Team impact

Removed the single point of failure

Previously only the original author could run the analysis. It now runs from a documented command with sample data included, usable by the whole team.

Project overview

Hardening a working prototype into a team tool.

The research group had a script that worked, but only in the hands of the person who wrote it, on their own machine, against one specific version of the dataset.

CIDS worked through the pipeline step by step: parallelising the slowest stage, adding input validation so a malformed file failed with a clear message instead of a silent wrong answer, and writing a short setup guide the rest of the team could follow without help.

The rebuilt tool now runs as part of the group’s regular workflow, and the original author has since used the freed-up time to start a follow-on study that wouldn’t have been possible while manually babysitting the old pipeline.

Researchers collaborating around a shared workstation. — The rebuilt pipeline runs on the same lab hardware as the original prototype - the goal was reliability and reach, not new infrastructure.20-minute runtime, down from 6 hours

Before and after

2 operational shifts from the project

Research-to-Production Bridge

Before

Who can run it

Only the original author, on their own machine.

Processing time

A six-hour single-threaded run.

After

Who can run it

Any team member, following a documented setup.

The analysis no longer depends on one person's availability.

Processing time

Twenty minutes with a parallelised pipeline.

Iteration on the analysis itself became practical again.