DTU.AI

• 

  login page

• 

  interpreter

• 

  example

Inspiration

Every DTU student struggles with finding and understanding college circulars. Most are buried in Telegram channels, uploaded as PDFs, or unclear images. Important details often get missed. The idea was to make something that reads these circulars automatically, summarizes them, and tells exactly who it’s meant for.

What it does

DTU.AI automatically fetches circulars from DTU Telegram channels, extracts text from PDFs or images, and uses the Gemini API to interpret them.
It gives:

• A short summary of the circular
  
• The branches or departments it applies to
  
• The type of notice (Exam, Holiday, Result, General)
  

So students don’t have to go through every message — they get clear, structured information instantly.

How we built it

The app is built with Python and Streamlit.
It uses:

• pdfplumber and pytesseract for extracting text from PDFs and images
  
• Google Gemini 2.0 Flash for summarizing and classifying circulars
  
• Telegram API for fetching documents automatically
  
• dotenv for API key management
  

A basic login screen was added for demo purposes to show a complete product flow.

Challenges we ran into

Setting up the Gemini API and managing environment variables took time.
PDFs and image circulars had very different formats, so OCR extraction needed fine-tuning.
Telegram’s document fetching system had some rate limits and required retries.
Python dependency conflicts and environment setup also caused multiple rebuilds.

Accomplishments that we're proud of

We made a working pipeline that connects Telegram to AI and Streamlit in one flow.
The text extraction and Gemini interpretation work smoothly on local setup.
A simple login interface and clean layout make the tool easy to present.
It solves a real DTU problem in a way that actually saves time for students.

What we learned

We learned how to connect APIs, manage OCR and NLP in one system, and keep the code modular.
We also got hands-on experience with Gemini integration, Streamlit UI design, and handling environment issues.
Most importantly, we learned how to turn a college inconvenience into a usable product.

What's next for DTU.AI

We plan to deploy it online with a live Telegram connection for real-time circular updates.
Add filters so students see only notices relevant to their branch or year.
Push mobile notifications when new circulars arrive.
Later, expand the platform to other colleges facing similar notice distribution issues.

Built With

• dontev
• google-gemini-api
• pdfplumber
• pillow
• pytesseract
• python
• streamlit
• telegram-bot-api