Resume Classifier API

📌 About the Project

The Resume Classifier API is a backend service designed to classify resumes (CVs) using machine learning models. It provides an API endpoint for uploading resumes as raw text or files (PDF), processes them to extract relevant content, and returns predicted job categories along with confidence scores.

This project is built using:

• ⚙️ FastAPI for the API layer
• 🧠 Langchain with Gemini LLM for resume analyze
• 📦 MongoDB for storing extracted resumes, classification results, and submission metadata
• ✅ Pydantic for input validation
• 🧪 Pytest for API and service-level testing
• 🐳 Docker for containerized development and deployment

The goal is to create a modular, scalable, and production-ready API that demonstrates backend and ML engineering skills while solving a realistic use case.

---

Demo

Users can submit their resume via POSTrequest to the /upload_resume/ endpoint. The system performs the following steps:

1. Validates file MIME type and it content
2. Extracts file content using the text_extractor_service.py
3. Analyzes the resume against the job description using Gemini via Langchain
4. Return a structured output:

• The classification (Match, Partial Match, No Match)
• A score between 0 and 1
• An AI generated tip with a practical advice of how to improve the resume for the job

Example Output

---

🗓️ Dev Log

[2025-06-15] - Project Kickoff

• Researched:
  • How to build RESTful APIs using FastAPI
  • Pydantic models for request validation and OpenAPI documentation
  • Connecting MongoDB with FastAPI using the async driver motor
• Started project scaffolding:
  • Created initial folder structure:
    • app/models for Pydantic schemas
    • app/db for MongoDB connection logic
    • app/services for ML and file parsing logic
    • app/api for route definitions
• Planned initial API route /upload_resume/:
  • Accepts uploaded file (UploadFile)
  • Extracts text and classifies using a discriminative model
  • Stores classification result in MongoDB

[2025-06-16] - upload_resume logic and Text Extraction Structure

• Defined the core /upload_resume route logic in routes.py:
• Began implementing the Text Extraction module in /services/text_extractor.py
  • Implemented specialized subclasses:
    • PdfTextExtractor()
    • MSWordTextExtractor()
    • PlainTextExtractor()
  • I've also created a centralized extractor selection via call_text_extractor()

[2025-06-19] - upload_resume logic and Text Extraction Structure

• Finished text_extractormodule and created some tests for it at ./tests/text_extractor.py
• I'm considering just using a LLM API for the classifier (although IDK if it would be the best decision)

[2025-06-21] - Persistency Layer, Docker Compose and API

• More Project structure organizing and renamed some folders
  • Added a Makefile for managing development commands.
  • Created a run.sh script to start both the API and MongoDB via Docker Compose.
• MongoDB integration:
  • Created a MongoDB conncetion handler DBConnectionHandler
  • Built a ResumesRepository with a working insert_document method for persistence.
• Created a docker-compose.ymlfor MongoDB startup
• Integrated lifespan context manager for startup and shutdown events:
  • On startup: establish DB connection
  • On shutdown: closes DB connection

  (I think it's a good one, i don't have anybody to ask for lol)

[2025-06-22] - AI Integration

• Implemented Resume Analyzer with LLM:
  • Introduced the use of Langchain with Gemini to process resume and job description inputs using Natural Language
  • Created a llm_classifier.pyand config.pyunder app/services/AI/to encapsulate prompt configuration and model interaction logic.
• Integrated AI analysis directly into the `/upload_resume/ route.
• Schema Updates:
  • Added job_description and ai_tipto the ResumeUploadRequestmodel.
  • Updated ClassificationDict with AI-generated suggestions for candidates.
• Cleanup and Refactor:
  • Removed obsolete classifier.py.
  • Updated requirements.txt to reflect Langchain and Gemini dependencies.