Ex-Servicemen Job Discovery Agent (Prototype)

Overview

The Ex-Servicemen Job Discovery Agent is a prototype AI-powered tool designed to assist ex-servicemen in India—particularly Junior Commissioned Officers (JCOs) and Other Ranks (ORs)—in discovering relevant civilian job opportunities. Built using the Google Agent Development Kit (ADK), the agent dynamically scrapes job postings from curated job boards, filters them based on relevance and recency, and formats the output in a structured markdown table.

It also includes an optional feature that allows users to generate and send a job summary via email, powered by the Brevo API.

---

🎯 Problem Statement Alignment

This prototype is developed in response to Problem Statement No. 6 (2025) issued by the Army Design Bureau, titled:

"AI Search Engine for Employment Opportunities of Ex-Servicemen."

📄 The complete set of problem statements is available in this official PDF.
📂 A dedicated copy of this specific problem statement is also provided in the problem_statement/ folder of this repository.

⚙️ Current Focus — Part-I
The prototype specifically targets Part-I of the deliverables:

✅ "AI-based aggregator application to search for jobs on the Internet specifically for Ex-Servicemen."

This aligns with the recognized need:

"Manual search of vacancy is time consuming and chances of missing out on good vacancies are high."

---

Features

• Job Scraping: Scrapes job postings from multiple job boards tailored for ex-servicemen using Firecrawl.
• Dynamic Filtering: Filters job postings to include only those from the current month, extracting relevant fields like organization, job description, and apply links.
• Structured Output: Presents job postings in a markdown table format with fields: Sl. No, Name of Organisation, Govt/PSU/Pvt, Description, Date, Location, Apply Link, and Other Relevant Links.
• Email Integration: Allows users to generate and send the job summary via email using Brevo, with a user-driven confirmation workflow.
• Modular Agent Design: Employs a multi-agent architecture with sub-agents dedicated to scraping, filtering, drafting, and emailing.

Project Structure

parent_folder/
  job_discovery_agent/
    ├── __init__.py
    ├── agent.py              # Root agent definition
    ├── prompt.py             # Instructions for the root agent
    ├── websites.py           # List of websites to scrape
    ├── .env                 # API keys for Firecrawl and Brevo
    ├── tools/
    │   ├── __init__.py
    │   ├── brevo.py         # Tool for sending emails via Brevo
    │   ├── firecrawler.py   # Tool for scraping websites using Firecrawl
    │   └── get_websites.py  # Tool to retrieve the list of websites
    └── sub_agents/
        ├── scrapy/
        │   ├── __init__.py
        │   ├── agent.py     # Scrapy sub-agent for scraping
        │   └── prompt.py    # Instructions for scrapy sub-agent
        ├── filter/
        │   ├── __init__.py
        │   ├── agent.py     # Filter sub-agent for processing scraped data
        │   └── prompt.py    # Instructions for filter sub-agent
        ├── writer/
        │   ├── __init__.py
        │   ├── agent.py     # Writer sub-agent for drafting emails
        │   └── prompt.py    # Instructions for writer sub-agent
        └── brevo/
            ├── __init__.py
            ├── agent.py     # Brevo sub-agent for sending emails
            └── prompt.py    # Instructions for brevo sub-agent

Prerequisites

• Python 3.9+
• Google ADK (pip install google-adk)
• Firecrawl (pip install firecrawl) for web scraping
• Brevo API (pip install sib-api-v3-sdk) for email functionality
• A local IDE (e.g., VS Code, PyCharm) with terminal access

Setup Instructions

1. Clone the Repository:

   git clone https://github.com/yourusername/ex-servicemen-job-discovery-agent.git
   cd google-adk-job-agent

2. Create a Virtual Environment:

   python -m venv .venv

   Activate it:

   • macOS/Linux: source .venv/bin/activate
   • Windows: .venv\Scripts\activate

3. Install Dependencies:

   pip install google-adk firecrawl sib-api-v3-sdk

4. Configure API Keys:

   • Open the .env file in the job_discovery_agent/ directory.

   • Add your Firecrawl and Brevo API keys:

     FIRECRAWL_API_KEY=your_actual_firecrawl_api_key_here
     BREVO_API_KEY=your_actual_brevo_api_key_here
     

5. Run the Agent:

   • Navigate to the parent directory:

     cd parent_folder

   • Launch the ADK development UI:

     adk web

   • Open the provided URL (e.g., http://localhost:8000) in your browser to interact with the agent.

Usage

1. Generate Job Summary:

   • In the ADK UI, send the command: generate.
   • The agent will scrape job postings, filter them for the current month, and display a markdown table with the following columns:
     • Sl. No
     • Name of Organisation
     • Govt/PSU/Pvt
     • Some Description about Post
     • Date
     • Location
     • Apply Link
     • Other Relevant Links

   Example output:

   | Sl. No | Name of Organisation | Govt/ PSU/Pvt | Some Description about Post        | Date       | Location    | Apply Link                | Other Relevant Links       |
   |--------|----------------------|---------------|------------------------------------|------------|-------------|---------------------------|----------------------------|
   | 1      | ABC Corp            | Pvt           | Software Engineer                 | 2025-05-01 | Remote      | [Apply](http://abc.com)  | [Details](http://abc.com/info) |
   | 2      | XYZ Govt            | Govt          | Administrative Officer            | 2025-05-05 | Delhi       | [Apply](http://xyz.gov)  |                            |
   

2. Generate and Send Email:

   • After the table is displayed, the agent will prompt: "Would you like to generate an email with this job summary? Reply 'generate email' to proceed."
   • Reply with generate email to draft an email.
   • The agent will show the drafted email and ask: "Here is the drafted email. Reply 'send email' to send it, or 'cancel' to stop."
   • Reply with send email to send the email via Brevo, or cancel to abort.
   • Screenshot of the same is being shown below:
   • 

Current Limitations (Prototype Stage)

• ❌ Parsing Accuracy: The agent relies on the LLM to parse markdown content, which may lead to inconsistent field extraction depending on the website structure.
• 📉 Website Coverage: Limited to a predefined list of websites; more job boards can be added.
• 📅 Date Filtering: Assumes dates are in a recognizable format; may miss jobs if dates are formatted unusually.
• 📧 Email Functionality: Requires a valid Brevo API key and assumes a default sender email.

Future Improvements

• 🧠 Enhanced Matching: Add NLP-based job matching to user profiles.
• 🔁 Dynamic Searching: Expand to dynamic location-based filters
• 🌐 User Interface: Introduce user interface for real-time interaction.
• 📄 Resume matching: Support resume matching and CV generation (Part-II of the problem)
• 🚨 Error Handling: Add robust error handling for failed scrapes or email sends.

Contact

For questions or feedback, please open an issue on GitHub or reach out to the maintainers.

---

Last Updated: May 18, 2025