March 2026 Update: Intelligent LinkedIn Enrichment, Custom Criteria Weighting & Multi-Tab Navigation

March has been a big month for ResumeRank. Three major features ship today, all driven by real feedback from our users — independent recruiters, staffing agencies, and in-house HR teams. Here's what changed and why it matters.

1. LinkedIn Enrichment: An Intelligent End-to-End Pipeline

LinkedIn enrichment already existed, but it relied on detecting a URL in the resume. No LinkedIn link in the CV? No enrichment. That's no longer the case.

What Changed

The new pipeline works in 4 automatic stages:

Stage 1 — Smart Extraction from the Resume Gemini 2.5 Flash analyzes the resume text and extracts the candidate's name, recent experiences, and — if present — the LinkedIn URL. No more fragile regex patterns: the LLM understands varied formats (URLs split across lines, partial links, etc.).

Stage 2 — Automatic Web Search If no LinkedIn URL is found in the resume, ResumeRank runs a targeted web search via Gemini 2.5 Flash + Exa search engine, filtered exclusively to linkedin.com. The query uses the candidate's name and most recent company to maximize accuracy.

Stage 3 — Profile Scraping Once the URL is identified, Apify extracts public profile data: work history, validated skills, manager recommendations, certifications, and language test scores.

Stage 4 — AI Validation This is the key addition: before integrating LinkedIn data into the analysis, a final Gemini 2.5 Flash call semantically compares the resume and the scraped profile. If the name doesn't match, if work histories diverge too much, the profile is automatically rejected. No more wrong LinkedIn profiles polluting your analysis.

Real Results

From our latest test batch (9 resumes):

The false positive rate dropped from ~15% to 0% thanks to AI validation. Cases where a wrong profile would have been scraped and integrated — like mismatched names or empty profiles — are now caught automatically.

Full Traceability

Every analysis now stores in the database:

• linkedin_url: the profile URL found (or null)
• linkedin_source: the source (cv or perplexity-search)
• linkedin_pipeline_trace: complete pipeline detail for debugging

This matters for agencies that need full traceability in their hiring processes.

2. Custom Scoring Criteria Weights

Until now, the 5 evaluation criteria (experience, skills, education, languages, soft skills) had fixed default weights. The problem: a senior developer role and a junior sales role don't have the same priorities.

What Changed

You can now customize the weight of each criterion directly from job creation or from the job management page.

Two adjustment modes:

1. Sliders — Drag to visually adjust each criterion's percentage
2. Direct input — Click the displayed percentage and type the exact value (e.g., 35%)

Validation logic:

• When using direct input, other criteria don't auto-adjust — you keep full control
• A warning message appears if the sum isn't 100% (e.g., "Total is 85%. 15% is missing — adjust criteria to reach 100%.")
• The Save button is disabled until the total reaches 100%

Practical Example

For a QA Tester with AI focus role:

By increasing technical skills weight to 40%, candidates with genuine testing/AI expertise naturally rise in the rankings — even if their academic background is less conventional.

3. Multi-Tab Navigation for Candidate Comparison

A simple change that transforms the daily experience: you can now open multiple candidate profiles simultaneously via middle-click or Ctrl+click.

Before

Click a candidate → read → back button → click another → read → back button. Tedious.

After

Ctrl+click 5 candidates → 5 tabs open → freely compare profiles. Natural.

Technically, candidate cards now use Next.js <Link> components instead of JavaScript onClick handlers, restoring native browser behaviors (middle-click, Ctrl+click, right-click > "Open in new tab").

4. Under-the-Hood Improvements

Less visible but equally important:

• Updated analysis model: Gemini 3.1 Flash Lite for paid accounts (via OpenRouter)
• Complete FR/EN translations: All new features are fully available in both French and English
• Better batch reliability: Processing 50+ resumes at once is more stable, with fewer "lost" CVs
• Localized error messages: Error messages in job creation, analysis page, and inline loader are now properly translated

What This Means for You

These improvements are available immediately for all users on the Founder's Offer plan. LinkedIn enrichment requires a paid account; criteria weighting and navigation are accessible to everyone.

Ready to try these features? Test them on your next recruitment or reach out to us for a personalized walkthrough.