May 7, 2026
6 min read
Build Notes

No Degree. Twenty Years of Receipts

What self-taught actually means when the classroom was a showroom floor, the curriculum was an agency you owned, and every claim has an artifact behind it.

Michael Donovan
Michael DonovanAI Engineer · Founder · Automotive AI Platform Builder
No Degree. Twenty Years of Receipts
Most dealers don't have an AI problem. They have a visibility problem.Vendors are happy to sell ten dashboards that never talk to each other. I have sat in your chair. I know which numbers move the needle and which ones just move invoices.The Signal is where I write down what actually works, what is vendor theater, and the plays I would run in your store this quarter. No buzzword salad. Just the field notes of someone who has carried a bag and shipped the code.

In 2005 I was a technician in a friend's independent shop. Motorcycles and cars, detailing, the tire machine, brakes, and rotations. No degree then. No degree now. Twenty years later I ship production AI platforms, and the most common question I get from hiring managers and fellow builders is some version of how. The honest answer is not romantic. It is a list of receipts, and the list is the point.

If you came up the traditional way, this is what the other path actually looks like from the inside. If you came up like me, this is your permission slip to stop apologizing for it.

Self-taught does not mean untrained

Self-taught gets heard as nights with YouTube tutorials and a certificate of completion nobody checks. That is not what happened. Self-taught means the training was real but nobody handed me a syllabus. I had to build the syllabus myself, in public, with money on the line, for twenty years straight.

The curriculum had three phases: the showroom floor as user research, the agency decade as product school, and certifications as structured checkpoints to verify the self-built map against the official territory. Each phase produced artifacts. That matters more than where the learning came from, because artifacts can be checked and stories cannot.

The floor was the user research

DOM360 hired me as its first employee, a Sr. Account Manager at an automotive digital-marketing startup. First hire at a startup means you do everything: project management, product, vendor management, paid and organic search, analytics. There is no better forced education than being the only person standing between the promise and the delivery.

Then Lazare Auto Group, 2009 to 2011. I sold cars and ranked among the top-50 US automotive salespeople in OEM rankings. Selling on a showroom floor is user research with consequences. You learn what buyers actually do, not what surveys say they do, and you learn it in real time with your paycheck attached.

Promoted to Marketing Director, I built a lean team that 4X'd lead volume and set up vendor-independent lead-source attribution using early Google Analytics plus cookies. Why vendor-independent? Because vendor-reported numbers are vendors grading their own homework, and I wanted the real answer.

Lia Auto Group came next, 2011 to 2012, as Director of Search Marketing. We brought SEO and SEM in-house across 18 dealership sites: a 34% total traffic increase, with organic accounting for 84% of traffic. I sat on a panel at Digital Dealer 11. None of this was a coding bootcamp. All of it was learning how systems, incentives, and measurement behave under pressure, which turns out to be most of engineering.

The agency decade was product school

In 2012 I co-founded OOMDO and ran it until 2023. Eleven years, 20+ employees, $60M+ in client sales over the decade, and proprietary lead-capture and conversion tech we built ourselves because nothing on the market did what our clients needed.

The flagship result: we took a Honda dealer from #182 to Top 25 nationally, from 150 to 500+ units per month, with three consecutive record months above $1.2M in variable operations. That kind of move does not come from a media budget. It comes from technology, process, and measurement working as one system, rebuilt and retuned for years.

An agency you own is the most brutal product school available. You design the tech, sell it, support it at nine at night when it breaks, and eat the consequences when it underperforms. There is no professor to appeal to and no curve. The market grades, and it grades in churn. Ten years of that teaches you more about software that survives contact with real users than any classroom could have taught me in the same window.

Certifications are checkpoints, not the credential

In 2024 I started adding structured checkpoints: Google Digital Marketing and E-commerce, Google AI Essentials, and Google Data Analytics Foundations through Coursera, then the Adobe Content Creator Specialization across 2024 and 2025. I am an active member of the Google Developer Program.

Notice the order. The certifications came after nearly two decades of operating, not before. They are not the foundation, they are verification: a deliberate audit of the self-built map against the official territory, patching the gaps that solo learning always leaves. I also keep earned and in-progress certifications strictly separate on every surface, because the moment those blur, every other claim on the page gets discounted with them.

The work kept compounding alongside the checkpoints. National Sales Director at Dealerverse from 2024 to 2025. VP of Marketing at Strolid, where monthly opportunities grew from an average of 12 to 50+ rooftops, MRR rose 40%, Q2 2025 set a company record, and I built an internal meeting-intelligence pipeline in TypeScript plus Python. Now Vandoko, the company I founded, where I am principal AI engineer and vandoko.ai is in live beta.

The real red flag is not the missing degree

Here is what twenty years of hiring people, vetting vendors, and being vetted myself has taught me: nobody competent rejects a candidate for lacking a degree. They reject candidates whose story does not hold together.

Vetting today is cross-examination by humans and AI at the same time. Your site, resume, LinkedIn, and GitHub get compared line by line, and the machines do not get tired. The red flags are gaps with no explanation, titles that inflate between surfaces, and numbers that drift depending on the audience. Inconsistent history reads as carelessness or fabrication, and both are disqualifying.

That is why I am precise to the point of stubbornness. I am credited with helping generate a reported $2.4B in dealer profit across my career. OOMDO did $60M+ in client sales over a decade. Two different numbers, two different scopes, never conflated, on any surface, ever. The discipline is not modesty. It is durability. Claims that survive an audit are the only claims worth making.

Receipts beat credentials

A degree certifies that you could probably learn. Receipts certify that you did, repeatedly, under load, with money on the line.

Top-50 OEM-ranked salesperson. Lead volume 4X'd with attribution I could defend. Eighteen sites brought in-house with organic at 84% of traffic. A decade-long agency with $60M+ in client sales. A dealer carried from #182 to Top 25. A 40% MRR lift. A meeting-intelligence pipeline in production. A platform in live beta. That is the stack, and none of it requires you to take my word for anything.

If you are self-taught and second-guessing yourself: stop optimizing for permission and start producing artifacts. If you are vetting someone like me: do not ask where they studied. Ask what they shipped, then check whether the story matches everywhere it is told.

See the receipts

The full record is on this site. The work itself is at /work, and if you want twenty years of pattern recognition pointed at your problem, the engagements are at /pricing.