The client
Series A fintech in San Francisco. Payments infrastructure. 35 people. $18M raised.
The product worked. The business worked.
What didn't exist yet was AI.
The CEO decided AI was the next core capability. Not a chatbot on the dashboard. Not a feature. A function.
Fraud detection, risk scoring, automated compliance — with its own person, its own roadmap, and eventually its own team.
Why the client came to us
The internal HR team was strong at closing engineering hires. But this wasn't a standard engineering hire.
AI talent sits at the intersection of research, engineering, and business — a space where traditional job boards and keyword-matching don't work.
The CEO needed a partner who understood the difference between someone who talks about AI and someone who builds it. That's where we came in.
The “beautiful unicorn” problem
The brief was for one person who simultaneously:
- understands business and strategy,
- can find real AI use cases — not “let’s bolt on an LLM,”
- can embed AI into product and processes,
- speaks with engineers on the same level,
- goes hands-on when needed,
- then builds a small team around the function.
This is what we call the “beautiful unicorn” problem.
People with “AI” in their headline are everywhere.
People who can actually build an AI function inside a business — much fewer.
Title matters
The client wanted “Head of AI.”
We pushed back.
At 35 people, you don't need a VP. You need a founding engineer who writes code on Tuesday and presents to the board on Thursday.
Title inflation attracts the wrong candidates. “Head of AI” brings in directors looking for a comfortable perch. “Founding AI Engineer” brings in builders looking for a problem.
What we did
1. Role recalibration
Reframed the position from “Head of AI” to “Founding AI Engineer.” Same scope, completely different signal. Aligned the brief with the actual stage of the company — zero-to-one, not scale.
2. Practitioner search
Filtered for shipped production models, not Kaggle medals or conference papers. Looked for commercial ML with real money flowing through it. Ignored proofs of concept that died in staging.
Production ML in a commercial context is a different skill from academic ML. You can tell in the first ten minutes of a conversation who has it.
3. Business judgment screening
One rule: the first AI use case must move a number. Revenue up or cost down. Not “demonstrate technical sophistication.” Candidates who started with infrastructure before outcomes — filtered out. Candidates who started with “here's what I'd measure” — stayed in.
4. Technical deep-dives
For the shortlist, we went deep: architecture decisions they'd made, failure modes they'd lived through, and what they'd do differently now.
Not a stress test. A reality check — had they actually been in the room when the hard decisions got made?
What this person actually does
This wasn't just an ML engineer. This was a one-person AI department.
Fraud & Risk
Build fraud detection models, risk scoring pipelines, real-time inference for payment transactions.
Compliance Automation
Automate AML checks and compliance workflows that were eating 20+ hours per week of manual work.
Infrastructure
Set up ML pipeline from scratch — data ingestion, feature stores, model training, monitoring. No existing ML infra to build on.
Team Building
Hire the first 2-3 ML engineers. Define technical direction. Build the function that didn't exist.
CEO-level Communication
Report directly to CEO on AI strategy. Translate technical capabilities into business outcomes. Present to the board.
Process
5 weeks
Time to close
14
Profiles shared
4
Interviews
1
Finalist
Selected candidate with:
- real experience building an ML pipeline from scratch at a mid-stage startup,
- production-ready models in a commercial context,
- startup mindset — no team when she arrived, three engineers when she left.
Result
Within 6 months:
- Fraud detection shipped. Chargebacks down 40%.
- 2 ML engineers hired.
- Reporting directly to CEO on AI strategy.
The function that didn't exist became a core differentiator in half a year.
“She moves like a founder.”
— CEO, after the first 90 days
Business impact
The company got more than an employee.
They got a strategic capability that didn't exist 6 months earlier:
- Fraud losses cut by 40% — directly hitting the bottom line
- AI became a competitive differentiator in investor conversations
- The hire attracted two strong ML engineers who wanted to work with her
- Time-to-compliance on AML checks dropped from days to hours
What this case is really about
Brief calibration. Not sourcing, not screening, not closing. The brief.
- Stopped the client from hiring the wrong profile. “Head of AI” at 35 people is a red flag.
- Recalibrated the brief before starting the search.
- Found a practitioner, not a title-holder.
- The function that didn’t exist became a core differentiator in 6 months.
The AI talent market is the noisiest market in tech right now. The gap isn't supply. It's signal. And the signal starts with what you call the role, how you describe it, and what kind of person that description attracts.