Projects / Recruitment
Shortlists in hours, not weeks
Recruitment Recruitment client
✦ Result
4 h
from posting to a ranked shortlist
An assistant that reads every application against the role, ranks them with reasons, and hands recruiters a shortlist they trust.
Before: 11 days from posting to shortlist
The problem
Popular roles drew hundreds of applications. Recruiters skimmed, good people were missed, and the shortlist took almost two weeks.
What we built
An assistant that reads every CV and cover letter against the role’s requirements, scores each one and writes two lines on why. Recruiters see a ranked list, open any candidate, and see exactly what matched.
Results
Shortlists arrive the same day a role closes. Recruiters spend their time talking to candidates, not sorting them.
How it was built
- ATS API
- CV parsing
- LLM scoring with reasons
- Bias checks
- Recruiter dashboard
- Evals on past hires
- Postgres
- Week 1: score past roles against real outcomes
- Weeks 2–3: scoring, reasons, bias checks
- Week 4: recruiters test on live roles
- Week 5: live across all roles
- 1 lead engineer
- 1 full-stack engineer
- Head of talent (client side)
FAQs
Does it reject people?
No. It ranks and explains. Recruiters decide who moves on.
How do you keep it fair?
It scores against the role's written requirements only, and we test every release for skew across groups.
Does it work with our ATS?
We built on their ATS. Most ATSs have an API we can use the same way.
Got a process like this? 20 minutes, and you'll know if it's worth building.
Book a call