Job Application Tracker — an AI automation workflow using n8n

## TLDR;
Built a local AI-based job application tracker using n8n + LLMs + Google Sheets.
- Fetches emails daily
- Filters only job-related ones
- Uses LLMs to classify and extract relevant data
- Appends results into a Google Sheet

All done locally — private, automated, and efficient.

Table of contents

Problem

What is the issue I had, and what was I trying to achieve then. My problem is that I put in a lot of job applications day in, day out, but I never had a way to properly track them. And I didn’t want to manually create an Excel/Sheets/Airtable entry every time I applied.

P.S. - As the complete setup is local to your system/network. Hence more private, preserving data privacy too.

So, what did I do then? I built an AI automation workflow that runs in the background — quietly, without interrupting my flow — and logs my job applications in Google Sheets.

Solution

Prerequisites

• Local n8n instance deployed via Docker (since the cloud version has only a 14-day trial and why share your private info anyway?)
• A local LLM running in LM Studio (or other) with exposed API. You can use OpenAI API too, if privacy isn’t a concern.
• Google Cloud account with Gmail, Drive, and Sheets APIs enabled, and OAuth credentials configured in n8n.
• Local LLM API connected in n8n as a credentialed resource.
• Wanna know about the setup I had for this? Check it out at the end.

Workflow

How It Works

1. Triggering the Flow: The automation runs every midnight via a schedule trigger, or manually via a button if needed.
2. Fetching Emails: All emails from that day are fetched using the Gmail integration.
3. Filtering:
   • Job listing platforms are excluded to avoid spammy alerts.
   • Subject/message content is checked for inclusion or exclusion keywords like “interview”, “thank you”, or “job alert”.
4. Text Classification: Emails that pass filtering are classified using a local LLM (Gemma via LM Studio) to determine if they relate to a job application.
5. Information Extraction: The LLM is used again to extract relevant details such as company name, job title, application status, and so on.
6. Appending to Sheets: If classified as a job application, the extracted data is appended as a new row in the connected Google Sheet.

All of this runs quietly and locally, and the sheet updates itself daily — no manual input required.

Output

Just blurred out the entries, but they do exist — don’t worry about that 😉.

With the entries, now I can do even more things like how many applications that I submitted in a day, in a month, and many more statistics.

Finale

Overall, this workflow works great for capturing application entries. I just do a quick review weekly to correct any occasional errors. Below are some limitations,

• Simplified email content sometimes causes the LLM to misclassify or miss context.
• The system doesn’t track duplicates or updates yet, as there’s no stable unique key per row. That’s a work-in-progress.

My Setup

• Had a n8n docker container with docker volume created and complete n8n setup done - running on my homelab's ubuntu server.
• Probably thinking of moving this container to my kubernetes cluster with a separate namespace and PVC for its data.
• Btw, for local LLM inference, I’m using LM Studio server for now with Google’s Gemma 13b parameter model on my RTX 2060 MAX-Q laptop.
• Really, a laptop? I’m broke and job hunting, not want to spend a dime more than what’s absolutely necessary.