← All postsBlog

Meet Marcus - the resume screening agent that runs on your laptop

· Avery NXR

Resume screening is one of those workflows that everyone agrees should be partly automated and that almost nobody automates well. The off-the-shelf ATS tools do a rough keyword match and call it done. The cloud-LLM tools work better but they send candidate PII to third-party AI providers, which is a posture that legal and compliance teams in regulated industries can't easily approve.

Marcus is one of the 7 production-ready agent templates that ship with Avery NXR. He monitors a folder for resumes, scores each candidate against the job description, classifies fit level, and drafts screening emails for the strong matches. He does all of it on your laptop. The candidate's name, email, address, and work history never cross to a third-party AI provider.

What Marcus actually does

You drop resumes into a folder (or hook him up to a jobs@ inbox, or to your ATS via webhook). Marcus then:

  • Detects the new resume
  • Extracts structured fields — name, contact, education, work history, certifications, skills
  • Compares the extracted profile against the job description you've configured
  • Classifies fit (strong / moderate / weak / not a fit)
  • Surfaces reasoning — what matched, what didn't, what's notable
  • For strong matches: drafts a personalized screening invitation email
  • For other categories: logs structured data and optionally drafts a polite rejection note
  • Records everything to your ATS or a structured store

The whole loop takes 30-90 seconds per resume.

Why PII concentration matters

Resumes are concentrated PII. Every resume contains the candidate's full name, contact information, home location, education history with dates that imply age, work history with employer names and tenure, and frequently more — immigration status, sometimes age, sometimes details that suggest ethnicity, religion, or other protected characteristics.

For some categories of role, the data is explicitly regulated. Healthcare professionals' resumes contain identifiers that interact with healthcare regulations. Defense workers' resumes contain clearance information. Government workers' resumes touch reporting requirements.

Sending all of this data to a third-party cloud LLM is a posture that:

  • Most legal teams object to once they think about it
  • The EU AI Act and state-level US legislation are increasingly restricting
  • Carries reputational risk if a candidate ever discovers it

Marcus moves all of this in-house. The data stays on your laptop. The processing stays on your laptop. Your compliance posture stays defensible.

What's running under the hood

Marcus's graph in Avery NXR:

Folder watcher / email trigger / webhook
  → New resume detected
  → Parse PDF/DOCX (file ops node)
  → Extract structured fields (local LLM)
  → Compare against JD (local LLM with rubric)
  → Classify fit level (classification node)
  → IF strong match:
       → Draft screening email (local LLM)
       → Send via Gmail/SendGrid
  → IF other:
       → Log structured data
       → Optionally draft polite rejection
  → Write to ATS or database

What it costs

A cloud-LLM equivalent costs $0.05-$0.15 per resume depending on length and the depth of analysis. For a high-volume recruiting function processing thousands of resumes per month, this is in the low-five-figures per year. For an enterprise TA function handling 100K+ resumes per year, the bill is in the six figures.

Marcus runs on your local model. Cost per resume is electricity.

The bias-audit story

Marcus has a specific advantage when it comes to bias auditing. Because the model is local and you can record every input, every output, and every decision rationale to an audit ledger, you can run bias-audit reports against your own data without third-party dependency.

This matters because the regulatory environment around AI in hiring is tightening fast. Several US states require bias audits. The EU AI Act classifies AI in employment as high-risk. Demonstrating fairness to regulators is much easier when the model, the data, and the audit trail are all yours.

Try Marcus in 5 minutes

If you've already got Avery NXR:

  1. Open the Agents tab
  2. Find Marcus (MARCUS · RECRUITING)
  3. Click "Use this template"
  4. Configure the source (folder, inbox, or ATS webhook)
  5. Paste in the JD you're screening against
  6. Plug in your email provider for the screening invitations
  7. Hit Run

Drop a test resume into the folder. Marcus produces the structured assessment and the draft email in about a minute.

If you don't have Avery NXR yet, request access at avery.software. Free Desktop tier, no card required.