Follow the curiosity. It knows where it's going.
Your quirks and differences aren't obstacles — they're the competitive advantage. Nobody handed me a map into AI engineering. I followed the thing I couldn't put down, and it turned out to be a career.
the printer is real. so is the curiosity.
People ask me how I became an AI engineer. The honest answer is that I never planned it — I followed a question I couldn't put down, and then another one, and they kept getting more expensive to answer.
I came up through consulting, where the job is to walk into someone else's mess and work out which part is actually broken. That training never left. When large language models arrived, plenty of people saw a demo. I saw a system with no guarantees — and I wanted to know exactly where it would fail.
So I started building. In Ruby, mostly, because it's the language I love and because plenty of people said it couldn't be done there. Retrieval pipelines. Agents with real tools. And eventually the unglamorous part nobody posts about: the evaluations that tell you whether any of it works.
- Test Double → Identus Consulting → DealMaker
- Ruby Central program committee
- RubyAI community — creator and admin
The turns that mattered
none of them looked like a plan at the time-
The spark
Blame Iron Man, honestly
The first time I wanted AI to exist, it was because Tony Stark had Jarvis and I wanted to know what a life with one would feel like. Years later I audited an AI class after my CS degree and walked away unconvinced — the nondeterministic systems of 2016 were less interesting to me than shipping software with my community. So I left AI alone.
the itch didn't go anywhere though
-
January 2023
I sat down at 5 PM and closed the laptop at 4 AM
ChatGPT had landed and I couldn't stop poking at it. By 4 AM one thought was completely clear: everything is going to change. The next morning I was ranting to my family, running on fear, excitement, sleep deprivation, and a little grandiosity. Software had been moving in a straight line and was about to bend.
I could not go back to work on Monday and pretend
-
The bet
Why Ruby? Because I was curious
Everyone was building AI in Python. I wanted to know what Ruby actually couldn't do — and I had a hunch that porting the ideas over would raise AI literacy in a community I love. Two niches, combined on purpose: it was a career repositioning and a curiosity itch at the same time.
combining niches is underrated
-
The work behind it
A certificate, not a masters, and a lot of evenings
Coursera felt too loose; Georgia Tech felt too long. UT Austin's certificate program had projects, deadlines, and real classroom pressure — exactly the structure I needed to get building sooner. Meanwhile I reproduced Python projects in Ruby and demoed what I found at company all-hands until I became the AI person people came to.
teaching it is how you learn it
-
RailsConf 2023
The room was full and people were lined up at the door
I pitched one talk to two conferences and got into both. Then came the hard part: writing it. Weeks of evenings and weekends turning a weather-prediction model into pure Ruby. When I walked in, the room was packed — people were catching the recording instead. That validated the whole bet in about ninety seconds.
first talk I ever gave. no pressure.
-
The pivot
RAG, because hallucinations were brutal
GPT-3.5 had a 4K context window and a habit of making things up. RAG was mostly an architecture problem, and architecture is language-agnostic — so I built the prototypes that let other Rubyists see themselves doing it too. Then I took it to Madison Ruby.
4K context. we were all doing hacky things.
-
On my own
Founder & AI Engineer — three clients and a lot to learn
I started a consultancy: AI strategy, custom implementations, training, and the parts nobody warns you about — marketing, sales, positioning technical expertise as business value. Hardest work of my career. Also some of the most rewarding.
running the business teaches you what the work is worth
-
Now
Dual-wielding Ruby and Python at DealMaker
Back in an IC seat, building AI applications for a platform helping companies raise over $2B. Reliability, retrieval, agents, and the evals that keep them honest. I couldn't have planned this out if I'd tried.
go figure.
What I believe
argue with me, I like itReliability is a design decision, not a model upgrade.
The next model will not save a system whose success was never defined.
Most eval problems are product problems wearing a lab coat.
If three people on your team grade a run differently, you have a spec problem, not a scoring problem.
The demo is not the system.
Anything works once. Reliability is what survives the hundredth run and the weird user.
Ruby was never the limitation.
Language choice is rarely why AI products fail. Unclear thinking is.
Curiosity compounds faster than credentials.
I got here by following questions, not a syllabus. It is a repeatable strategy.
Explain it plainly or you don't understand it yet.
Jargon is where confusion hides. If I can't explain the architecture to a non-engineer, I go back.
If you're making the same turn
what I'd tell you over coffee-
Invest your time strategically
A resume is fine. Being able to name the problems an organization is about to face is better.
-
Build in public
Writing, sharing the deep work, showing the messy middle — that's what put founders and recruiters in my inbox.
-
Go deep in a few places
You don't need to know everything. You need a few areas down tight, tied to real use cases.
-
Talk about real problems you solved
Not the tools you touched. The hallucination you chased down and what finally fixed it.
-
Be open about what you don't know yet
Naming the gap reads as expertise. Pretending there isn't one never does.
Still following it, honestly.
These days curiosity looks like evals, agent architectures, Ruby tooling that shouldn't exist yet, and an ongoing argument with myself about how much autonomy is too much. If any of that is your problem too, I'd like to hear about it.