Auditable transparent resume screener

If you need a resume screener that is auditable, explainable, and does not use black-box neural networks, I'd focus on a rules-based screening engine rather than a typical AI recruiting product.

The key distinction is that a tool can be transparent about its scores while still using neural networks under the hood. You want to be able to inspect the actual screening logic, reproduce a result, and explain exactly why a candidate received a particular score.

Best options to investigate

• 

1. HireLens

(Best starting point)

Open source · Self-hostable · Configurable scoring

HireLens provides weighted rubrics, evidence-linked scores, versioned criteria, hash-chained audit logs, blind review, and adverse-impact reporting. It can run with a deterministic heuristic fallback, but also supports LLMs and semantic embeddings. You must explicitly configure and verify a non-neural scoring path to meet your requirement.github.com

Links: GitHub repository ·Live demo

---

• 

2. ResumeParser

Open source · Offline · Lightweight

A good building block if you want straightforward, inspectable parsing and skills matching rather than AI-generated candidate judgments. It runs locally and shows how resume content is interpreted. It is more of a parser and skills checker than a complete, auditable hiring decision system.github.com

Link:GitHub repository

---

• 

3. Qcandidate

Commercial software · Evidence-linked scoring

Qcandidate advertises configurable scoring criteria and evidence attached to each candidate score. However, its public description does not establish that its underlying scoring is free of neural networks. Treat it as a vendor to evaluate, not a confirmed match.qcandidate.com

Link:Qcandidate website

What I would choose

For a strict no-neural-network requirement, I would start with HireLens as a prototype, or build a deterministic screening engine around an inspectable parser if you need stronger guarantees.

A transparent rules-based screener could work like this:

1. Parse resumes

Extract text and fields using conventional parsing rules, with no neural OCR or AI-based extraction.

2. Apply explicit criteria

Required qualifications, verified skills, relevant experience, and predefined keyword or phrase matching.

3. Calculate a reproducible score

Use documented weights and deterministic arithmetic. No embeddings, neural classifiers, or LLM judgments.

4. Produce an audit record

Store the rule-set version, matched resume evidence, score calculation, reviewer actions, and final decision rationale.

What to demand from any vendor

• Written confirmation that neural networks are not used in parsing, matching, ranking, or scoring—not just that generative AI is optional.
• Inspectability: documented rules, configurable weights, and access to the implementation or sufficient technical documentation.
• Reproducibility: the same resume and rule-set version produce the same score.
• Evidence traceability: each score can be traced to specific resume text and a specific criterion.
• Fairness testing: measure selection rates and investigate disparate impact; transparent rules alone do not guarantee fair outcomes.
• Human oversight: scores inform recruiter review rather than automatically rejecting candidates.

One important caveat: conventional OCR can itself use neural networks, even if the scoring engine is entirely rules-based. Verify the complete pipeline.

One question so I can narrow this down

Are you looking for a ready-to-use product, or are you open to self-hosting or building a rules-based screener? That determines whether I should prioritize commercial vendors with contractual guarantees or inspectable open-source implementations.