From Building Prototypes to Building AI

Graffiti Books and Break Dancing

Growing up in LA, I was the artsy kid — but in a street way. I was a break dancer. I was a graffiti artist. Not on walls — I never tagged anything real — but my sketch books were filled with it. Wildstyle lettering, characters, the whole thing. That was what cool looked like in LA in the '90s.

As I got older, the art shifted to cars. I was deep in the import tuner scene. I had a 1994 Honda Accord, four-door, stick shift. I wanted to fix it up but like every teenager, I was broke. I did what I could with what I had — working minimum wage jobs and putting every spare dollar into that car.

After high school I had no plan. No college acceptance letter. No family tradition of higher education. What I had was a thing I was interested in: cars. So I went into auto body.

And something clicked.

The Process: Seven Careers, One Thread

Auto Body — Learning to See What's Actually Broken

Auto body wasn't just painting cars. I got obsessed with it. I painted my own Accord. I shaved the door handles, added body kits, learned to lay fiberglass. I started making molds for body kits — actual molds, not just bolting on aftermarket parts. I was fabricating.

That's where I learned to diagnose problems by tracing symptoms back to root causes. A door doesn't close right? Is the frame bent? Did the impact shift something three panels over? You can't just fix the obvious damage — you have to trace the force through the whole system.

That's debugging. I didn't know the word yet, but that's what I was doing. It's the same skill I use today when a production API returns 500 errors and the logs look clean. Something upstream shifted. You trace it back.

Art Center and the Design Pivot

One of my friends went to Art Center College of Design for high school and introduced me to the college. I walked in and fell in love. Industrial design. Transportation design. These people were designing the cars I was modifying in my garage.

I took a couple of classes there and realized it was incredibly hard to get accepted into the full program. So I pivoted — enrolled at the Art Institute of California for media art and animation. I enjoyed it, but my heart was in industrial design. I dropped out.

Art school was expensive and I wasn't convinced I'd make money doing art. But I'd taken a model making class at Art Center that changed everything. We learned wood working, clay modeling, and I was already good at fiberglass and painting from auto body. I was a natural at it.

My first idea was becoming a clay modeler, but those jobs were rare and the industry was changing. So I set my sights on model making, and the logic was simple: if I couldn't design cars as a career, I could learn design on my own — but if I knew how to build them, I'd always have work. And if I ever did design something, I'd be the rare person who could actually make it real. I always enjoyed the real thing over just a design. A rendering on a screen is cool. Holding the actual part in your hands — that's the thing.

That logic turned out to be right in ways I didn't expect. Today, if I really wanted to design and build my own car, I could. Every step of it. That was the bet, and it paid off. The same logic applies to software and AI now — I don't just design systems, I build them. End to end.

But back then, the problem was practical: there was no dedicated program for model making. No school taught "how to be an R&D prototype model maker."

ARRK — The Real Education

One of my teachers was building a model making program in Orange County, and I started taking classes with him. He connected me with a job at ARRK Corporation in San Diego — one of the oldest Japanese model making companies in the world, founded in 1948 in Osaka.

That's where the real education started. CNC programming, 3D printing, mold making, fiberglass, painting, assembly. I was learning every step of the prototype pipeline, touching every material, running every machine.

Then something happened that changed the trajectory of my career.

A new hire came in straight out of school. A manager position opened up — a role I was qualified for, a role I'd been doing informally. He got the job. Because he had a degree. I was the one teaching him the ropes. He was my student. But on paper, he outranked me.

That stung. Bad. I decided to go get my degree. The Art Institute of Orange County had just opened an industrial design program, so I enrolled. It was expensive, but I wanted that credential. After a few quarters, I could see the gap. I'd taken classes at Art Center — one class there was worth three or four at Art Institute. The depth, the rigor, the expectations. Once you've experienced that level of education, you can't unsee the difference. I was paying Art Center prices for a fraction of the education. I dropped out again.

That's twice now. Two schools, no degree. I went back to model making.

5 Axis Design — Concept Cars and 28-Hour Days

I got a job at 5 Axis Design, and that's where things got serious.

We built full drivable concept cars and show models. I'm talking about chopping cars in half and rebuilding everything. Making full car molds, building parts from scratch, assembling entire vehicles. The craftsmanship standards were insane — everything had to be show-quality.

And the timelines were brutal. I remember working a 28-hour work day. Twenty-eight hours straight. And here's the kicker — you don't get triple time after 24 hours. The clock just resets. That was the culture. Short deadlines, no excuses, deliver.

That's where I learned what craftsmanship under pressure actually looks like. Quality AND speed. Not one or the other. That lesson shows up in every sprint I've ever worked.

Xerox — Thick Skin and a Reality Check

Then the 2008 recession hit and I got laid off from 5 Axis. Model making jobs dried up. I needed income, so I tried something completely different: Xerox. Business-to-business copier sales. Not phone calls — I was physically canvassing businesses. Walking into offices cold, trying to get past the front desk, pitching to people who didn't want to talk to me.

Sales taught me persistence and thick skin in a way that building things never did. But the sales cycle for expensive copiers was brutal — six to eight months or longer before a deal closed. I had a kid by then and I wasn't making enough. I couldn't wait half a year for a commission check when there were diapers to buy. I had to go back and get a real job.

But I kept the thick skin. That turned out to be worth more than any commission.

Honda and the End of Model Making

After Xerox, I went back to what I knew — small prototype shops, 3D printing shops, and eventually landed a mechanical engineering job at a blow mold shop — a three-person operation where I did SolidWorks CAD and Mastercam CNC programming.

Then I got the opportunity to work at Honda R&D Americas in Torrance as a contractor.

The thing about me is I was so interested in all of it that I learned everything. Design, fabrication, welding, metal work, wood, hand modeling, clay modeling, 3D CAD/CAM. If you gave me a napkin sketch — just a concept — I could figure out the design and build it as a prototype. Any material, any process.

Over the course of my model making career — a full decade across ARRK, 5 Axis, small shops, and Honda — I was building future vehicle designs. Things that wouldn't be on the road for years. Taking anything from napkin sketches to full engineering drawings and turning them into physical objects you could hold, test, iterate on.

Here's what I didn't realize at the time: that's spec-driven development. Someone hands you requirements — sometimes clear, sometimes contradictory, sometimes incomplete — and you figure out how to make it real. That's literally what software engineering is. The material is just different.

But being a contractor at Honda meant I was capped. No raises. No benefits. No path to full-time. It's a Japanese company — someone had to die for a headcount to open up. No one ever left. No one ever got fired. I was stuck.

I had to make a decision: go back to school for a mechanical engineering degree, or try something completely new.

The Leap — Coding Bootcamp with Two Kids

I chose software. It was more promising, and I was already a mechanical engineer in practice — I didn't need the paper to know I could think that way.

I started at community college for prerequisites while teaching myself to code on the side. Web development. HTML, CSS, JavaScript. Then I noticed this new thing: coding bootcamps. Dev Bootcamp had just launched maybe a year earlier. It was brand new.

I saw an opportunity. I started learning Ruby on Rails and found a bootcamp to attend.

But there was a problem. I was the sole provider for my family. Two kids. A wife. I would have to quit my job to do the program full-time.

I talked to every software engineer I knew. They said yeah, it could work. So I made an executive decision. I took what savings I had, put it into the program, quit my job, and told myself: if this doesn't work, I'll hustle. I used to do Xerox sales — I'm not shy about knocking on doors.

That was the scariest decision of my life. Not because I doubted I could learn it — I'd taught myself harder things. But because this time, it wasn't just me. Two kids were eating dinner every night because of a bet I made on myself.

It worked.

Zurb and Beyond

After the bootcamp, I landed a job at Zurb — a distinguished design firm that built Foundation, one of the most popular CSS frameworks at the time. Zurb was the real deal — a product design company that built tools used by developers everywhere. Bootstrap's creator, Mark Otto, had actually worked at Zurb before going to Twitter, where he built Bootstrap. The two frameworks were rivals, but Zurb was where a lot of that thinking originated.

From there, it was multiple jobs across multiple stacks — Ruby on Rails, React, Node, GraphQL, AWS. Every role taught me something different. Government contracting at VA.gov taught me rigor. Startups taught me speed. The Washington Post taught me scale.

But the transition was never as hard as people expected. Because I already knew how to solve problems that hadn't been solved yet. I already knew how to be resourceful. One of the biggest lessons from R&D was: fail fast. Get to the hardest problem as quickly as you can so you have time to solve it. That's not a software principle. That's a model making principle. I just brought it with me.

What Transfers

Here's the thing nobody tells you about career changes: nothing is wasted.

Break dancing and graffiti taught me creativity and pattern recognition. Design thinking before I knew the term.
Auto body taught me to diagnose root causes, not just symptoms. That's debugging.
Model making and clay sculpting taught me to build from specs, iterate fast, and work with my hands. That's prototyping.
CNC machining taught me to program machines to execute my design with precision. Read that again: program machines to execute your design. That's AI-augmented engineering.
Working 28-hour days at 5 Axis taught me craftsmanship under pressure. That's shipping production software on a deadline.
A decade of model making taught me spec-driven development before the term existed. Napkin sketch → design → prototype → test → iterate.
Xerox sales taught me thick skin and persistence. The sales cycle for expensive copiers was 6-8 months or longer — you learn to keep showing up when nothing's happening. That's every long debugging session and every slow-moving enterprise deal.
Quitting my job with two kids taught me that conviction is a skill, not a feeling.

The people who are best at AI engineering right now aren't just the ones who understand transformers and token windows. They're the ones who know how to think about systems, design for constraints, and iterate when the first approach doesn't work. That's not a CS skill. That's a builder skill. And I've been building things my whole life.

What It Actually Felt Like

I want to be honest about the hard parts.

Every transition was terrifying. You go from being the expert — the person people come to with questions — to being the beginner who doesn't know the vocabulary. Your identity shifts. You're not "the model maker" anymore. You're "the new guy." Again.

Watching someone with a degree get the job you earned at ARRK — that stays with you. Dropping out of school twice because the education wasn't worth the debt — that feels like failure even when it's the right call. Sitting at your kitchen table at 2am trying to understand JavaScript closures while your kids are asleep — that doesn't feel like a career transition. That feels like gambling with your family's future.

"Do it till it happens" — that's my philosophy. But I want you to know what that phrase sounds like at 2am when you're failing. It doesn't sound motivational. It sounds desperate. It sounds like: I don't have a plan B, so this has to work.

And it did work. Every time. Not because I'm special or talented or gifted. Because I kept showing up. Because the same stubbornness that kept me at the CNC machine at 11pm until the part came out right is the same stubbornness that kept me at the keyboard until the code worked.

There is no ceiling. The ceiling is me.

Still Building

I spent a decade building prototypes of cars that didn't exist yet. Physical objects made from concepts and specs, cut on machines I programmed myself, iterated until they met a standard most people would never see.

I'm still doing the exact same thing. The specs are product requirements now. The machine is an AI model. The material is code instead of aluminum. But the work — understanding the system, designing the approach, building it, testing it, making it better — that hasn't changed at all.

If you're reading this and you're thinking about making a change — maybe you're a mechanic curious about code, a machinist hearing about AI, or you're in sales and you miss building things — your skills are real. The thinking transfers. The gap between where you are and where you want to be is smaller than it looks.

It just doesn't feel that way at 2am.

The tools change. The thinking doesn't.

Here's what excites me about where we're headed: the future with AI is more accessible now for the creative person who persists and learns. Intelligence is being augmented by AI — but the builders, the creatives, the people with ideas and the stubbornness to execute them — they're the ones in control. Small teams can produce what used to take departments. One person with a vision and the right tools can build what used to require a company.

If you have an idea and you do it till it happens, the path is wider now than it's ever been. I had to change careers six times and teach myself everything the hard way. The next person with that same drive won't have to. AI is the great equalizer for builders.

I'm still building things that don't exist yet. The material just changed.

— Bill John Tran