Guy Walks Into a Radiologist's Office

I was finishing up my Cone Beam Computed Tomography (CBCT) Observation rotation. The radiologist and I had just finished reviewing five cases together. Our conversation drifted to the use of AI in dentistry. We talked about how good these new models seemed to be at pattern recognition, and how naturally that capability mapped onto CBCT diagnosis and reporting. Could a model surface differential diagnoses from a DICOM volume? Could it draft the report a radiologist would otherwise spend an hour writing?

At the time, I was trying to start an indie iOS/macOS software business. I had a demo of my current project on my phone, and on a whim, I pulled it out and showed him.

He gave me a look of awe.

I want to be honest about how much that small moment mattered. I was a dental student showing a half-finished side project to a senior clinician. Instead of polite acknowledgement, I got genuine surprise. It was a look that said, you can actually build things. That look lodged itself somewhere in my brain. It was the spark. It made me feel like our conversation wasn’t just idle speculation. It was something I could, in principle, actually go build.

I left the office that day with a different idea of what my career might look like.

In the months that followed, I went deeper. The technical hill was steeper than I had hoped, but the more I learnt, the more convinced I became that there was real opportunity here for a company, and not just a side project.

By the time 2024 rolled around, I had started talking to every dentist who would listen. What did they hate about their workflow? Where did AI actually fit, and where was it a solution in search of a problem? I still wanted to pursue CBCT diagnosis, but the deeper I got, the more I saw the intermediary mountains: ethically acquiring training data, navigating regulatory approval, building clinician trust, and so on. None of these are reasons not to do it. They’re just reasons it can’t be your first move.

So I pivoted to something more tractable: a GenAI assisted dental referral program. I used this idea to apply to Y Combinator under the name Anthrodontics. While waiting for the results, I moved to Emerald, Queensland, to start my first job. Most new graduates take a metro job, ease in, and enjoy their evenings. I went bush, treated patients all day, and tried to build a startup at night.

In May 2025, I relocated to Sydney to pursue the idea more seriously. After a year of ruminating, attending dental AI seminars and conferences, and drawing on experience from the industry, I think I finally got it.

It’s been four years since I first started imagining this future, and it has been, by every honest measure, a rollercoaster. I’ve had quite possibly the most unconventional new grad experience of anyone in my cohort: moving twice in less than a year, moonlighting as a self-taught programmer,, trying to imagine what the future of dental AI will look like.

Would I do it again?

Probably not.

The romantic answer would be of course, in a heartbeat. But the truthful answer is that the cost has been real. Friendships lost, weekends evaporated, a clinical career that has taken a back seat to a startup that may or may not work. The look of awe from that radiologist set me on a path I didn’t fully understand the price of.

So here’s my parting advice, equal parts sincere and tongue-in-cheek: the next time someone gives you a look of awe and respect, ignore the call.

You probably won’t. I didn’t.

But at least you’ll have been warned.