HR · internal hiring

Referrals screened before a recruiter opens them

HR posts a role with its job description. Employees refer candidates with a CV. Before anyone reads either document, the system compares them and returns a structured verdict — score, skill match, experience gaps, risk flags and a written reason.

5 stages

referral pipeline with email at every step

15+ fields

in the structured verdict, not a paragraph

In daily use

by an internal HR team

ORVELA

Shortlist

Employee referrals

HR

Candidate

Referred by Meera T.

RoleSenior Frontend Engineer
Referred14 Jun 2026
StageUnder review
CVcandidate-resume.pdf
JDsenior-frontend-jd.pdf
78

Strong match

Confidence: high

Automated screen

A human reads every verdict — this ranks, it does not decide

15 fields

Skill match — 71%

React TypeScript Node.js PostgreSQL Next.js AWS Kubernetes

Critical gap: none

Experience

Candidate years6.2
Required years5.0
Meets requirementYes
Gap months0
⚑ Career break detected — 7 months, 2023Overqualified: noOverlapping roles: none

Written reason

Matches five of seven required skills including the full core stack, and exceeds the experience bar by 1.2 years. Cloud and orchestration experience is the gap. The 2023 career break is flagged for context, not as a negative — it is stated on the CV.

Swipe to see the full screen

The verdict is a structure, not a paragraph — so a recruiter can sort and filter by it.

The situation

Employee referrals are the best hiring channel most companies have and the worst-managed. A CV arrives as an email attachment, sits in an inbox, and the referring employee never learns what happened.

The screening itself is the bottleneck: someone has to read a job description and a résumé side by side and form a judgement, for every referral, and the judgement is not written down anywhere comparable.

What I built

A referral portal with an automated first-pass screen, where the output is a structure rather than a paragraph.

  • HR posts a role with its job description attached; employees refer candidates with a CV.
  • Referrals move through submitted, under review, interview scheduled, and then rejected or hired — with email notification at every step, so the referring employee is never left wondering.
  • On submission, the job description and the résumé are compared automatically and a verdict is persisted against the referral.
  • The verdict carries a score, a confidence level, skill match percentage with matched, missing and critical-gap skills listed separately, candidate years against required years, gap months, career-break and overlap detection, an overqualified flag, risk flags, a recommendation, and a written explanation.
  • A reporting view over the referral funnel, with stage-by-stage conversion and time-in-stage.
ORVELA

Shortlist

Employee referrals

91 referrals

Submitted

34

Under review

18

Interview

11

Hired

6

Rejected

22

Referrals

Screened automatically on submission — sorted by match

CandidateRoleReferred byMatchStage
Ananya R.Senior FrontendPriya N.78Interview
Vikram S.Data AnalystArjun D.71Under review
Meera T.Store ManagerKavya M.66Under review
Rohit K.Backend EngineerNikhil P.54Submitted
Priya N.MerchandiserDivya B.49Submitted
Arjun D.QA EngineerSameer V.38Rejected

Swipe to see the full screen

Every referral’s stage, visible to the employee who made it.
ORVELA

Shortlist

Employee referrals

Last 90 days

Referrals

91

18%

Interview rate

31%

6%

Hire rate

6.6%

1.1%

Avg time to interview

4.2 days

1.8 days

Where referrals stall

Stage conversion, measured rather than guessed

Submitted → review
62%
Review → interview
48%
Interview → offer
27%
Offer → hired
84%

Swipe to see the full screen

Where referrals stall, measured rather than guessed.

Decisions worth explaining

The choice, why it was made, and what it means for you. Skip the middle part if you like — the point still lands.

01

The output is a structure, not a paragraph

The choice

The screen returns fifteen-plus discrete fields rather than a block of prose.

Why

A recruiter can sort and filter a structure. A paragraph, they have to read — which means at volume they do not, and the screening adds nothing.

What it means for you

You can rank a week of referrals by skill-match percentage, or filter to everyone with a critical gap, without opening a single CV.

02

It screens, it does not decide

The choice

A human reads every verdict, and the flags exist to be argued with.

Why

An automated hiring decision is both a bad idea and, in most jurisdictions, a growing legal problem. The flags — career break, overqualification — are exactly the ones where an automated judgement would be wrong most often and most unfairly.

What it means for you

You get the time saving without handing over the decision, and the reasoning is written down so it can be challenged.

03

Company sign-in, not another password

The choice

Authentication runs through the company’s existing directory.

Why

An internal tool with its own user list becomes a stale user list the day someone leaves.

What it means for you

Access follows employment automatically, and there is no separate offboarding step to forget.

For the technical readerUnder the hood
  • Express 5 and TypeScript backend on MongoDB with Mongoose, directory-based single sign-on, and transactional email.
  • React and Vite front end with Redux Toolkit and persisted state.
  • Document parsing runs as a separate evaluation service, so the portal stays responsive while a CV is being read.

What it changed

  • Referrals stopped dying in an inbox.
  • The first screening pass takes minutes instead of days, and produces something comparable across candidates rather than an opinion per CV.

Where this applies to you

Any company running an internal referral scheme, or any high-volume hiring process where the first pass is a human reading two documents side by side. The same document-comparison shape applies well beyond hiring — tender responses against requirements, claims against policies.

Screenshots show the real interface with fabricated candidates and roles. No real applicant data appears anywhere on this site.

Tell me what’s slow, manual, or unreliable.

Twenty minutes on a call is usually enough for me to tell you whether software is the answer and roughly what it would cost. If it isn’t, I’ll say so.

Available for freelance builds alongside senior engineering work. One or two projects at a time.