Load the context
A daily schedule starts downloads of the base resume and cover letter from Drive. PDF text extraction prepares the reference material.
n8n × APIFY × OPENAI × GOOGLE WORKSPACE
Less job hunting.
More intention.
A workflow that turns scattered LinkedIn listings into a focused application queue—with fit assessments and tailored documents ready for a human review.
Try the interactive demo ↘✳
LINKEDIN JOB TRACKER
& APPLICATION TAILOR
Find. Assess.
Tailor. Review.
Built with n8nHuman in the loop ↗
3LinkedIn search streams
07:00Configured daily trigger*
70 / 100Application priority threshold
01 / THE PROBLEM
Job searching has
too many moving parts.
Discovering roles, spotting duplicates, comparing requirements, and rewriting application materials can fragment a job search. This project brings those steps into one repeatable n8n pipeline, with Google Sheets as the working queue.
The goal: keep the candidate focused on deciding where to apply and reviewing the evidence behind each recommendation.
02 / EXPLORE THE EXPERIENCE
Your morning shortlist.
Fictional jobs, scores, and document excerpts. Runs in your browser; no live LinkedIn, AI, or Google Sheets requests. No applications are submitted.
- 01 Check queue
- 02 Discover & dedupe
- 03 Assess fit
- 04 Tailor documents
- 05 Save for review
Ready to explore. Run the demo to process one sample job.
Sample jobs6
Ready for review0
Priority matches0
Marked applied0
03 / UNDER THE HOOD
One workflow.
A considered handoff.
Based on the supplied n8n export. The demo condenses the node graph into five readable stages.
WORKFLOW / PART 01
Open full image ↗From morning trigger to job queue.
Load the resume and cover letter, check ExistingJobs, and branch into three LinkedIn searches when the queue needs replenishing. Normalize the results, save new jobs, and pass a queued job into the application loop.
WORKFLOW / PART 02
Open full image ↗From AI assessment to review-ready documents.
Prepare the AI output, copy and fill separate Google Docs templates, clean up the resume, then merge both document branches. Write the application details to ExistingJobs and return to the loop after a wait.
Original workflow screenshots supplied for this project. Open either image for a closer view; node labels are limited by the resolution of the source captures. The interactive shortlist above uses fictional sample data.
Work the queue first
Read ExistingJobs. When no non-applied jobs remain, run three Apify searches: junior data, marketing analyst, and growth marketing in Greater Vancouver.
Normalize & deduplicate
Standardize fields, remove URL tracking parameters, and compare LinkedIn IDs and URLs with saved jobs. Keep up to 30 new records and select one queued job per run.
Assess, then tailor
Ask OpenAI for a structured fit assessment and tailored content grounded in the base resume. Copy Google Docs templates, replace fields, and clean up unused resume content.
Bring it back to Sheets
Append the score, rationale, document links, and “To review” status to ExistingJobs. A score of at least 70 sets an application-priority flag in the preparation step.
*07:00 is the trigger hour in the export; the active workflow timezone and successful scheduled execution have not been verified.
04 / DESIGN DECISIONS
Automation with
room for judgment.
Fit scores help prioritize attention. They are evidence-based estimates, not interview probabilities. Every generated application still needs a candidate’s review.
Evidence before keywords
The AI prompt restricts candidate claims to the base resume and treats job descriptions as reference data. This is a prompt constraint, not a guarantee of factual accuracy.
Weighted fit assessment
The prompt specifies experience (40), skills and tools (30), seniority (20), and education (10), with missing evidence treated as unverified.
Templates as the document layer
Google Docs copies provide the output structure; replacement and cleanup steps insert the tailored text before saving links to the tracker.
IMPLEMENTATION NOTES / CURRENT EXPORT
A working design, with clear next steps.
The exported “Select 5 Jobs” node uses slice(0, 1). “Update Existing Job” uses an append operation, so processing adds a review row rather than updating the original. Queue checks exclude only “applied” rows, which can allow a pending or review job to be selected again.
Before a production rollout: update rows by job ID, distinguish processed jobs from pending jobs, reconcile conflicting output-schema instructions, and validate complete runs. The export references ExistingJobs; it does not include a write to the separate Dashboard sheet. No live execution results or time-saving metrics are claimed here.
RESEARCH → WORKFLOW → REVIEW