In today’s fast-paced hiring environment, resume screening plays a critical role in identifying the best candidates efficiently. With companies receiving hundreds of applications for a single role, building a smart and reliable Applicant Tracking System (ATS) is no longer optional, it’s essential.

This guide walks you through a refined approach to creating a smoother, more intelligent end-to-end resume evaluation system that uses AI to analyze resumes against job descriptions—without the technical setup headaches.

 A Cleaner Solution: Direct PDF Text Extraction

Instead of converting PDFs to images, this improved system extracts text directly using PyPDF2, a lightweight Python library built specifically for reading PDF content. This update:

• Eliminates dependency issues
• Reduces processing time
• Improves accuracy and reliability

Workflow:

• Upload a resume in PDF format
• PyPDF2 reads each page and pulls the text
• The text is passed to an AI model for analysis
• Results are displayed in a user-friendly interface

 Quick Setup:

To get started, create a clean Python environment using conda and install the necessary libraries. Here’s a detailed chat:

API keys for the AI model should be securely stored using environment variables to protect sensitive information.

 Smarter Resume Analysis

The system uses Google Generative AI to evaluate resumes based on job descriptions. This is where the real power of the system lies—prompt engineering guides the AI to act like an experienced recruiter or ATS engine.

The AI is instructed to:

• Compare resume content to the job description
• Identify relevant and missing keywords
• Score the match percentage
• Provide tips for improving the resume

Example Prompt:

“Act as an experienced applicant tracking system. Analyze the resume against the job description. Provide a match score, list of missing keywords, and recommendations for improvement.”

 Python code Snippet:

1. Load Environment Variables & Configure Gemini

Securely load your Gemini API key using dotenv and initialize the Gemini SDK.

from dotenv import load_dotenv
import google.generativeai as genai
import os

# Load .env file
load_dotenv()

# Configure Gemini
genai.configure(api_key=os.getenv("GEMINI_API_KEY"))

2. Convert Resume PDF to Image for Gemini Input

Convert the first page of a PDF resume into a base64-encoded image for Gemini’s input.

import io
import base64
from PIL import Image
import pdf2image

def input_pdf_setup(uploaded_file):
    poppler_path = r"C:Program Files (x86)popplerLibrarybin"
    # Convert PDF to images
    images = pdf2image.convert_from_bytes(uploaded_file.read(), poppler_path=poppler_path)
    # Get the first page as image
    first_page = images[0]

    # Convert image to byte array
    img_byte_arr = io.BytesIO()
    first_page.save(img_byte_arr, format='JPEG')
    img_byte_arr = img_byte_arr.getvalue()

    # Return base64-encoded image data
    return [{
        "mime_type": "image/jpeg",
        "data": base64.b64encode(img_byte_arr).decode()
    }]

3. Prompting Gemini to Analyze the Resume

Use Gemini’s generative model to analyze the resume against a job description.

def get_gemini_response(prompt, pdf_content, job_description): 
      model = genai.GenerativeModel('gemini-1.5-flash') 
      response = model.generate_content([prompt, pdf_content[0], job_description]) 
return response.text

4. Streamlit UI for Upload & Input

Create a user-friendly interface to upload resumes and input job descriptions.

import streamlit as st
# Set page configuration
st.set_page_config(page_title="ATS Resume Expert")
# Page title and description
st.title("ATS Resume Analyzer")
st.markdown("Upload your resume and paste a job description to get instant AI feedback.")
# Job description input
input_text = st.text_area("Job Description:")

# File uploader for resume
uploaded_file = st.file_uploader("Upload Resume (PDF)", type=["pdf"])

# Upload success message
if uploaded_file:
     st.success("Resume uploaded successfully!")

5. Action Buttons for Different AI Functions

Trigger Gemini-powered resume evaluation, improvement suggestions, or ATS match scoring.

# Evaluate Resume button
if st.button("Evaluate Resume"):
     pdf_content = input_pdf_setup(uploaded_file)
     st.write(get_gemini_response(prompt_eval, pdf_content, input_text))

# Improve Resume button
if st.button("Improve Resume"):
     pdf_content = input_pdf_setup(uploaded_file)
     st.write(get_gemini_response(prompt_improve, pdf_content, input_text))

# Match Score button
if st.button("Match Score"):
     pdf_content = input_pdf_setup(uploaded_file)
     st.write(get_gemini_response(prompt_match, pdf_content, input_text))

6. AI Prompts Used for Gemini

Customize Gemini’s behavior using specialized prompts for evaluation, optimization, and matching.

prompt_eval = "You are an HR expert... [resume evaluation instructions]"
prompt_improve = "You are a resume consultant... [resume improvement instructions]" 
prompt_match = "You are an ATS scanner... [percentage match instructions]"

Result Screenshots:

Seamless Interface with Streamlit

A Streamlit web app makes this solution easy to use, even for non-technical users. The app allows you to:

• Upload PDF resumes
• Paste in a job description
• Instantly receive AI-generated feedback, including match score and keyword suggestions

No complex setup. Just a clean, functional interface.

Final Thoughts

In today’s automated hiring process, it’s important to be smart about how you stand out.

With this clean and modern approach to resume analysis, you can:

• Save time
• Improve screening accuracy
• Offer real insights backed by AI

It’s time to transform your resume screening process into something smarter—and simpler.

Neha Vittal Annam