Ehsan Naderi, Ph.D.

Product Design · Human Factors

I design things people use
when getting it wrong
has a cost

I find the hard problem in regulated medical devices, enterprise AI, and spatial research, then design for the case where a mistake is a consequential event, not a missed click.

About

I look for the problem first. The place where getting it wrong has a real cost, and I design for that case, not the happy path.

I started in industrial product design, where a bad decision ships in metal and plastic and can’t be patched later. That’s still how I think about software, including the years I spent designing for Google Cloud AI.

Before Google, I spent four years as a tenure-track Assistant Professor of Product Design at the University of Minnesota. The research habits stuck: I hold a Ph.D. from the University of Missouri-Columbia, my work on perception and presence is peer-reviewed, and I’m comfortable inside the rigor that regulated medical-device design demands.

Where the work began.

Before the regulated systems and enterprise platforms, there was the bench: several years of industrial design, prototyping, and research craft. It’s the foundation the rest is built on.

Where Getting It Wrong Has a Cost

Enterprise AI

Where confident output hides the quiet errors no one checks.

Medical Devices

Where use-error is a clinical event, not a support ticket.

Spatial Experiences

On how people actually interpret what they see.

Selected Work

Surgeons were arming a surgical foot pedal by accident mid-procedure, and couldn’t feel which speed mode they’d engaged.

I ran the pressure study that set the activation thresholds, then shaped the travel and a tactile detent so the surgeon feels each mode engage instead of watching for it. In summative testing, the accidental activations stopped appearing.

Analysts were trusting an AI’s confident answers and quietly giving up on checking the ones it got wrong.

I ran the pressure study that set the activation thresholds, then shaped the travel and a tactile detent so the surgeon feels each mode engage instead of watching for it. In summative testing, the accidental activations stopped appearing.

People who’d never written code were building real software — and the AI meant to help them kept guessing wrong against workflows they already depended on.

I designed the AI-assisted creation, automation, and database surfaces, where every change had to earn its place against what power users already relied on. It shipped as the first generative-AI feature out of Google Cloud; over the following months AI feature use rose ~400%, with 2,000+ apps built through it.

Radiologists were missing findings that were right in front of them — and AR overlays were sending field technicians to the wrong part.

I ran eye-tracking studies on where expert attention actually goes under pressure, then designed displays and overlays around the cues people used and dropped the ones they ignored. The same methods carried into a head-mounted mixed-reality study and a VR operating room. The work is peer-reviewed.

Contact

If the cost of getting it wrong is real on your side, that’s the conversation I want.