AI recruiting
AI recruiting for early-stage startups
Refery helps early-stage startups hire AI talent by separating the actual work from the title. We screen for shipped systems, evaluation judgment, production reliability, customer context, and the candidate’s personal contribution.
An illustration of how a Refery search is presented. The names and grades are examples, not real candidates.
Our clients are backed by investors including:
When this search fits
The roles Refery is built for.
Use this search when you need someone to ship an AI product, build an evaluation system, own inference or retrieval infrastructure, or deploy into customer environments. Research, product engineering, and infrastructure are treated as different hiring problems.
- Applied AI engineers
- Machine learning engineers
- Retrieval and inference engineers
- AI infrastructure engineers
- Forward deployed engineers
- Heads of AI
How Refery runs the search
From brief to a reasoned shortlist.
Name the production outcome
Start with the workflow, data, quality bar, latency, cost, and acceptable failure modes.
The brief- Must own
- The evaluation harness and the first model in production
- Stage
- VC-backed, Seed to Series B
- Still open
- Location, band, and the first milestone
Scope and fee agreed before the search15%Find evidence, not keywords
Search for people who can explain what shipped, how it was evaluated, and what failed.
The shortlist, with the reasoningEthan ChoCo-founder intro · work reviewedA+Ines CarvalhoSpecialist search · screenedA−Lena BrandtScout intro · screenedATest judgment
Use a bounded work discussion that covers baselines, release criteria, monitoring, and fallbacks.
What comes back- Evidence
- What they owned, and how we know
- Confirmed
- Interest, location, notice, comp expectation
- To test
- The one thing the interview should settle
An illustration of how a Refery search is presented. The names and grades are examples, not real candidates.
What the shortlist should prove
Evidence, not title matching.
- Task-specific evaluation experience
- Production software ownership
- Clear model-versus-software tradeoffs
- Evidence of reliability, latency, and cost decisions
Commercial terms
- Placement fee
- 15% to 25%of first-year base salary, paid only when you hire
- Retainer and exclusivity
- Nonerun Refery alongside your own channels
- Guarantee
- 90 daysunder the client agreement
The exact scope and fee are agreed before the search starts. See fees and terms →
From a real search
Calibrate on work, not an AI label
In a Refery engineering search, we grouped example profiles around different kinds of technical work instead of treating them as interchangeable. Each profile included a fit rationale, and candidate interest remained explicitly unconfirmed during calibration.
The client could tell us which work it wanted to explore. Later recommendations included the evidence behind our view and what the hiring team still needed to test. A relevant project was a reason for a conversation, not proof of every skill on the brief.
The useful lesson
For an AI search, name whether the immediate need is product delivery, research, infrastructure or customer deployment before comparing profiles.
Read the anonymized search story →Go deeper
Related searches and guides
Start a search
Tell us what this person needs to own.
Share the role, the outcome, and the constraints. Refery replies with how the search would run and what the shortlist should prove.