depa panjie purnama

Posted on Aug 29 • Edited on Aug 31

Your Job Finder Companion

#devchallenge #n8nbrightdatachallenge #ai #webdev

n8n and Bright Challenge: Unstoppable Workflow

This is a submission for the AI Agents Challenge powered by n8n and Bright Data

What I Built

I built Your Job Finder Companion, an intelligent AI agent that automatically checks LinkedIn, Indeed, and Glassdoor for new opportunities and delivers concise, AI-analyzed summaries via Telegram. The agent helps you catch fresh roles quickly so you can apply before listings get crowded.

Demo

Workflow Video:
Telegram Video: Only Available on YouTube Short
LinkedIn Screenshot:
Indeed Screenshot:
Glassdoor Screenshot:
No job listings available Screenshot:

How It Works

Send a Telegram message in the format:

Role name - Country

Examples: Developer - US or Data Scientist - UK

The agent launches snapshot queries on the three platforms using those inputs and returns up to one listing per platform. You may receive up to three messages (one per platform), each including an AI summary and a direct apply link when available.

Example LinkedIn response format:

🎯 LinkedIn Job Found
━━━━━━━━━━━━━━━━━━━━
🏢 Google
💼 Senior Software Developer
📍 Mountain View, CA
💰 $150,000 - $200,000

👥 Current Applicants: 12 candidates
📅 Posted: 2 hours ago

🤖 AI Analysis:
- SCORE: 92
- Excellent opportunity at a leading tech company...

🔗 Apply: [Direct Link]
━━━━━━━━━━━━━━━━━━━━

Or if no jobs are found:

❌ Sorry, no job listings available for the specified keyword within the last 24 hours.

n8n Workflow

The current workflow consists of 21 nodes orchestrating the entire job discovery and notification process.
The complete workflow JSON is available here: GitHub Gist - Your Job Finder Companion

Technical Implementation

System Architecture

Telegram Trigger: Receives user messages with job criteria
Message Parser: Extracts role and country from "role - country"
Data Collection: Creates Bright Data Marketplace Dataset snapshots for LinkedIn, Indeed, and Glassdoor
Snapshot Processing: Polls snapshot metadata until ready, then retrieves content
AI Analysis: Uses Google Gemini to score and summarize the listing
Notification Delivery: Sends formatted messages back to Telegram

Model and Tools

AI Model: Google Gemini via the n8n LangChain Google Gemini Chat Model node
Communication: Telegram Bot API
Data Processing: JavaScript Code nodes for parsing and shaping data
Error Handling: Wait/poll loops for snapshot building; platform-specific failure notifications

Bright Data Verified Node

The Bright Data integration is central to this solution, providing access to three marketplace datasets:

LinkedIn Dataset (gd_lpfll7v5hcqtkxl6l)

Filters: job_title includes role, country_code equals ISO-2 (uppercased), application_availability true, url includes http, job_posted_date within last 24 hours.
Example 24h filter:

{
  "name": "job_posted_date",
  "operator": ">",
  "value": "{{ new Date(Date.now() - 24 * 60 * 60 * 1000).toISOString() }}"
}

Indeed Dataset (gd_l4dx9j9sscpvs7no2)

Filters: job_title includes role, country (lowercased), date_posted equals "Just posted".
Example:

{ "name": "date_posted", "operator": "=", "value": "Just posted" }

Glassdoor Dataset (gd_lpfbbndm1xnopbrcr0)

Filters: job_title includes role (no recency filter in this version).

Journey

Development Process

Phase 1: Planning

Mapped user input to platform-specific filters and recency expectations
Researched Bright Data's marketplace datasets
Designed the workflow architecture

Phase 2: Implementation

Built n8n workflow with Telegram + Bright Data + AI analysis
Configured Bright Data nodes with appropriate filters
Implemented platform-specific recency constraints

Phase 3: AI Enhancement

Added Gemini agent to produce a SCORE and short prioritization summary
Developed scoring algorithm (1-100 scale)
Created platform-specific message formatting

Phase 4: Error Handling

Implemented waits/polling for snapshot readiness
Added failure messaging per platform
Created user-friendly error notifications

Challenges and Solutions

Snapshot Timing
- Different build times across datasets
- Solution: Poll metadata; if not ready, wait and retry until content is retrievable
Data Consistency
- Different schemas per platform
- Solution: Normalize essential fields and route by URL (LinkedIn/Indeed/Glassdoor)
Recency Alignment
- Varying recency controls per dataset
- Solution: LinkedIn uses last 24h; Indeed uses "Just posted"; Glassdoor matches title only in this version
User Experience
- Deliver concise, actionable messages
- Solution: Single best match per platform, AI SCORE + summary, direct apply link when possible

Key Learnings

Bright Data's marketplace datasets make it straightforward to access structured job listings without custom scraping
The snapshot pattern scales well for timely data retrieval
A compact AI summary helps users prioritize at a glance
Small UX choices (input format, per-platform messages) significantly impact usefulness
Proper error handling is crucial for user experience

Your Job Finder Companion demonstrates how n8n's workflow capabilities combined with Bright Data's robust datasets can create practical solutions that deliver immediate value to job seekers, ensuring they never miss opportunities from the last 24 hours.

Top comments (13)

Gary Xia • Aug 29

@depapp This is very cool! I wanted to use the Crunchbase/Pitchbook datasets, but saw that the marketplace datasets had a minimum order amount of $250. Was this the case for the 3 datasets you used as well? Any insight appreciated!

depa panjie purnama • Aug 29

@myopicoracle hey, thanks a lot
for my project I actually didn’t purchase the datasets directly from the Bright Data marketplace. Instead, I just used the Bright Data verified node in n8n and pulled them via the “snapshot by dataset” option. That way I could use the data right away without hitting the $250 minimum order thing.