CVsprings scores candidates with a transparent, rules-based engine. No language model ever reads your candidates' data. Same CV, same job profile, same result. Every time.
Built for boutique recruitment agencies in the Benelux & DACH
| Requirement | Result | Evidence from CV |
|---|---|---|
| Go | Matched | Led the technical design of the settlement platform, built in Golang. |
| Kafka | Matched | Designed Kafka event streaming between payment and ledger services. |
| Rust | Not found | No matching text in the CV |
Paste the job description, start from a template if you like, and set the weights for skills, experience, education and job title.
Upload one CV or a whole batch. Every CV is scored by the same rules, so the ranking is consistent.
Open any candidate to see the evidence behind every match, record your shortlist decision, and keep the audit log and bias report ready for your client.
Each assessment shows how the score is built: which skills from the job description the CV contains, next to the text that matched, and which it doesn't. When your client asks why, you answer in one click.
The same CV and job description, scored by the same engine version, always produce the same scores. Every report records the engine version and the weights it ran under. Try that with an LLM-based screener.
Every saved assessment is stored with the scores, weights and engine version behind it, and every later change is logged with who made it and when. Export the audit trail per candidate or per screening run.
No candidate data goes to third-party AI providers, because there are none in the pipeline. The CV file is discarded once it's scored, only the assessment record is kept, and per-candidate deletion is available on request.
A role and its weights, twelve CVs ranked, the top candidate's report with evidence for every skill, a shortlist decision, the audit log and the bias report.
Less magical.
Entirely defensible.
CVsprings, engineering principle nº 1
Most screening tools put a language model between your candidates and your shortlist. It reads the CV, forms an impression, and returns a score it cannot explain and will not repeat. Run the same CV twice and you can get two different answers.
Under the EU AI Act, screening candidates is high-risk when it's done by an AI system. Risk management, data governance, technical documentation, logging, human oversight, conformity assessment: a good part of that lands on you, the agency deploying the tool.
So we made the unfashionable choice: a rules engine you configure. You set the weights before you screen; they're recorded with every assessment and applied the same way to every candidate. No model reads the CV.
Every score comes with its working: the weights, the sub-scores and the evidence that matched. And a determinism check runs on every change: a fixed set of CVs must reproduce their recorded scores exactly, or the build doesn't ship.
Start free. Upgrade when you need unlimited analyses.
That trial link can’t be used any more — it may have expired or already been started. You can still start on any plan below, or book a demo and we’ll set one up for you.
Not sure which plan fits? Book a demo
Placeholder: one or two sentences from Jasper go here.
Founder, CVsprings · Made in the Netherlands · jasper@cvsprings.com
Send us 3 anonymised CVs and your job spec. In a 15-minute call we run them through the engine in front of you — twice, so you can watch the outputs match.
No slideshow, no generic sample data. You leave the call knowing exactly what CVsprings says about your candidates, and why.
Prefer to look before you talk? Start with our one-page EU AI Act checklist for recruitment tooling — no form needed.