Automated Resume Shortlisting (Batch Processor)

This workflow replaces manual resume screening with an AI-powered batch processor that:

• Pulls pending applications on a schedule
• Reads and converts resume files into text
• Evaluates each candidate against a DevRel job description
• Updates statuses to Shortlisted or Rejected and notifies the team on strong matches

Recruiters get consistent, fast evaluations instead of grinding through PDFs.

Previous State

New applications arrive → recruiter downloads resumes → skims each one → compares to job description → updates a spreadsheet or ATS → pings the team when someone looks promising.

The result: slow response times, inconsistent decisions, and a lot of time lost context-switching between tools.

Target State

• A time-based trigger wakes the workflow every 10 minutes
• It fetches a small batch of Pending candidates from the database
• Each resume is extracted, scored by AI, and classified as DevRel-fit or not
• Matching profiles are marked Shortlisted and pushed to Slack/Discord
• Non-matches are marked Rejected to keep the pipeline clean

The HR team only reviews curated shortlists instead of raw resumes.

Workflow Breakdown

1. Time Based Runs (Cron Trigger)

• Purpose: Start the batch automatically every 10 minutes.
• Cron: */10 * * * *
• Inputs: None – the workflow fetches its own work from the database.

This node is the heart of the batch cadence.

2. Query Data (Fetch Pending Applications)

• Collection: Applications
• Filter: status == "Pending"
• Limit: 5 (to keep batches small and avoid latency / throttling)
• Output: Up to 5 candidate records, each containing at least name, status, and resume_url.

If no pending records exist, the workflow naturally becomes a no-op for that run.

3. Loop Process (Process Each Candidate)

• Loop variable: {{nodes.query_data[]}} (array of pending applications).
• Purpose: Run the same inner steps (extract → evaluate → update → notify) once per candidate within the batch.

Downstream nodes reference the loop variable (e.g., {{loop.currentCandidate.resume_url}}) for candidate-specific data like the resume URL, name, and record ID.

4. Text Extraction (Read Resume File)

• Input: {{loop.currentItem.resume_url}} (PDF/DOC URL)
• Purpose: Convert the resume file into clean, machine-readable text.
• Output: content field containing the raw resume text.

This normalizes different file formats into a single text stream for AI.

5. Text Generation (AI Resume Evaluation)

• Model: aicon-v4-large
• Response type: JSON
• Prompt (simplified):
  • Provide the DevRel job description (community + content + coding, not pure backend).
  • Provide the resume text from the extraction node.
  • Ask the model to decide if the candidate is a DevRel fit and explain why.
• Expected JSON output:

• Role: This node is the core classifier: it turns unstructured resume text into a structured hiring decision.

6. If-Else Condition (Match vs. No Match)

• Condition: {{nodes.text_generation.output.is_match}} == true
• Branches:
  • IF true: Candidate is considered a DevRel match.
  • ELSE: Candidate does not fit the DevRel role.

This single decision point drives the rest of the pipeline.

7A. IF Match → Shortlist + Notify

Update Data (Shortlist)

• Collection: Applications
• Selector: _id == {{loop.currentItem._id}}
• Update fields:
  • status = "Shortlisted"
  • ai_notes = {{nodes.text_generation.output.reason}}

Send Discord / Slack Message (Alert)

• Channel: Recruiting / hiring channel (via webhook).
• Message example:

This ensures strong candidates surface to the team immediately.

7B. ELSE No Match → Reject

Update Data (Reject)

• Collection: Applications
• Selector: _id == {{loop.currentItem._id}}
• Update fields:
  • status = "Rejected"

Purpose: keep the pipeline clean by explicitly closing out mismatched profiles.

8. Return State (Batch Complete)

• Message: "Batch processing complete."
• Optional payload: Number of records processed, number shortlisted, number rejected.

This node makes the workflow easy to monitor from logs or higher-level orchestrators.

Conditional Logic & Branching

There is a single key decision:

• IF is_match == true → Shortlist + notify
• ELSE → Reject

No nested branching is required. All complexity remains in the AI evaluation prompt and the downstream updates.

Data Flow & Transformations

End-to-end flow:

1. Database → Query Data: Pulls up to 5 Pending applications.
2. Query Data → For Loop: Iterates through each candidate as loop.currentItem.
3. For Loop → Text Extraction: Downloads and extracts the resume file into plain text.
4. Text Extraction → Text Generation: AI maps text to a DevRel-fit decision + justification JSON.
5. Text Generation → If-Else: Routes candidates to Shortlist vs. Reject.
6. If-Else → Update Data: Writes back Shortlisted / Rejected and AI notes.
7. Update Data → Notifications: Sends alerts for shortlisted profiles.

Main transformation chain: File URL → Text → AI Classification → JSON Decision → Status update.

Integration Requirements

• AI model: aicon-v4-large for resume understanding and DevRel fit classification.
• File extraction engine: PDF/DOC ingestion for the Text Extraction node.
• Slack/Discord: Webhook URL or bot token for real-time notifications.

Authentication:

• AI API key (configured at the Text Generation node or globally).
• Webhook secrets / tokens for Slack or Discord.

Rate limiting:

• Batch size (limit = 5) keeps processing within reasonable time and under API quotas.
• Adjust the cron interval or limit upward/downward based on your inbound volume.

Outcomes

• Dramatically reduced manual resume screening for HR and hiring managers
• Consistent application of the DevRel job criteria across every candidate
• Faster response times for strong applicants
• Clear, auditable notes (ai_notes) explaining why someone was shortlisted or rejected

Recommended Extensions

• Add a “Maybe” bucket for borderline candidates with a separate follow-up workflow.
• Add an Email node to send automated acknowledgement or rejection messages.
• Store an AI score (0–100) alongside is_match for more granular ranking.
• Add an analytics dashboard view over statuses, match rates, and sources.

Next Steps

• Configure the Applications collection with at least name, status, and resume_url.
• Wire the Text Extraction and Text Generation nodes with your file store and AI credentials.
• Plug in Slack/Discord webhooks for shortlist alerts.
• Turn on the workflow and let it screen new applicants every 10 minutes.

👉 Install this template in WorqHat and let your AI batch-processor handle resume shortlisting.