Back to Projects

DocuSense: Intelligent PDF Management

An end-to-end project where I designed and engineered a smart document tool to turn digital clutter into an organized, queryable library using data science and machine learning.

Role:Product & ML Engineer
Duration:May 2023 (Self-Project)
Tech Stack:
Python
Pandas
Scikit-learn
NLTK
PyPDF2
Pytesseract
Jupyter
View Code on GitHub
A collage of the DocuSense UI, showing the dashboard, search, and clustering features.

1. The Challenge: Taming Digital Chaos

As a student, I was dealing with a growing collection of disorganized PDFs—lecture notes, research papers, and personal documents. Finding a specific file was often a frustrating, time-consuming process.

This "digital clutter" leads to inefficient search and wasted storage. I saw an opportunity to build a tool that doesn't just store PDFs, but understands them.

PDF scrolling demo
A messy closet representing digital clutter.

2. The Process: From Idea to Prototype

I approached this project with a dual mindset: as a Product Manager defining the "what" and "why," and as an Engineer building the "how."

Product Thinking & Design

I started by defining the user pain points, brainstorming core features, and designing a clean UI in PowerPoint to make the functionality feel simple and accessible.

Technical Implementation

I built the data analysis pipeline using Pandas and PyPDF2. To enable the core features, I implemented:

  • MD5 Hashing for efficient duplicate detection.
  • TF-IDF & Cosine Similarity to engineer a content-based search engine with Scikit-learn.
  • K-Means Clustering to automatically group similar documents based on their content vectors.
  • OCR Integration with Pytesseract to extract text from scanned PDFs.

3. The Solution: DocuSense in Action

The result is a practical PDF management tool built on a foundation of data science. The prototype delivers three core features:

DocuSense Dashboard

Analytics Dashboard

Instant insights on your entire PDF collection.

DocuSense Search

Content-Based Search

Find files by what's inside them, not just by name.

DocuSense Clustering

Intelligent Clustering

Automatically groups similar documents together.

4. Future Scope & Learnings

This project was a valuable exercise in building a solution from the ground up. It provided a solid foundation, and the clear next steps for expanding its capabilities would be:

  • Text Summarization: Integrate a transformer model to provide quick summaries.
  • Supervised Classification: Use user-generated labels to train a model for smarter, personalized tagging.
  • Cloud Storage Integration: Connect directly to Google Drive, Dropbox, and OneDrive.
Back to Projects