Blue Star NDT Platform

Non-Destructive Testing is a high-stakes field. Operators use software to run ultrasonic, radiographic, or magnetic inspections on industrial components — quality control for pipelines, aircraft parts, structural steel. The margin for error is essentially zero. Blue Star had a working product. But the complaints were consistent: training new operators took too long, experienced ones still made navigation errors, and the interface slowed down inspections that should have been faster. The ask wasn't a redesign for its own sake. It was about cutting the time between "start inspection" and "reliable result," with fewer mistakes in the gap. Key constraints from the start:

• No rewrite of the Visual C++ codebase
• Changes had to be technically feasible for the existing team
• Operators ranged from week-old trainees to 10+ year veterans
• NDT workflows vary meaningfully by inspection type

Watching operators work was the most valuable thing I did. People describe their workflow as cleaner than it actually is — they skip the parts that feel embarrassing, or they've simply forgotten what's hard because they've adapted to it. Sitting next to someone running an actual inspection tells you things an interview never will. One operator had a shortcut sequence she ran at the start of every shift — four clicks through menus that should have been one. She'd done it so many times she'd stopped noticing it was a problem. Her newer colleague, six months in, hadn't figured out the shortcut yet. He was still going the long way. That gap was the problem in miniature. Our analysis and research identified five core insights:

• Navigation Hierarchy Deficit: Tools needed constantly were as buried as tools used once a month. There was no structure based on frequency or phase.
• Mental Models Over UI: Operators were doing the software's job for it. Mental models filled in where the interface failed. When someone left or switched roles, that knowledge left with them.
• Canvas Conflict: Operators needed to focus on the inspection data — the waveform or image output. The workspace was cluttered with panels that pulled focus away from the actual work.
• Visual Noise: High contrast, minimal spacing, no clear visual hierarchy. The interface always felt urgent, even when nothing needed attention.
• Feasible Scope: Technical constraints were real, but narrow. The dev team couldn't rearchitect core modules, but the UI layer had room to move.

These came out of the research. I used them to anchor decisions and, honestly, to push back when things got political:

• Canvas first: The inspection data is the work. Everything else is in service of it.
• Visible by default, configurable on demand: Frequent controls should be reachable without thinking. Rare controls should exist — just not in the way.
• Reduce decisions, not capabilities: Operators shouldn't be making interface decisions mid-inspection. The interface should handle that.
• Respect the domain: NDT has conventions. Industrial software has visual patterns operators already know. Don't modernize at the cost of that familiarity.

The redesign was about reorganization and resurfacing — not reinvention. Ribbon navigation replaced the nested menu structure. Tools were grouped by inspection phase, not by when they'd been added to the codebase. Operators could see what was available for their current task without digging. Dedicated toolbars put the highest-frequency controls outside the menu system entirely. One click for common actions. Contextual settings panels separated settings that changed per-inspection from settings that almost never changed. Canvas-first layout gave the inspection workspace more room. Panels could collapse. Controls defaulted to positions that didn't compete with the data view. Low-contrast light theme reduced visual fatigue over long inspection sessions and matched conventions from other industrial tools operators used. The changes were incremental by design. We were making the product usable without memorization — not rebuilding it from scratch.

I tested prototypes with actual operators. Task completion exercises using real inspection scenarios, the same tasks they ran in the field. A few things changed: 1. Toolbar placement was wrong in the first pass. Operators expected certain tools on the left — consistent with other industrial tools they used. We moved them. 2. Contextual panels needed more explicit labels. Operators wanted to know which phase the settings applied to. 3. The ribbon had too many tabs. We consolidated based on how operators actually moved through an inspection, not how we'd assumed they would. The feedback cycle was tight. We weren't running six-week research rounds. Feedback came in, designs changed, we tested again. Getting the fundamentals right mattered more than getting them right on the first try.

Outcomes came from prototype testing and stakeholder review — the product was at design handoff, not production, so there are no live analytics to pull from. Client feedback was direct: the redesigned interface felt like it was built for operators, not for the engineers who built the underlying software. That's what I was going for.