CampusSaathi AI

CampusSaathi AI is a multilingual, low-bandwidth companion for Bharat’s students—helping them learn smarter, manage campus life, find opportunities, and support mental well-being in their own language

Comment

CampusSaathi AI: India-First Student Companion

Inspiration

India’s college ecosystem is incredibly diverse—students speak dozens of languages, study in varied infrastructure settings, and often face academic pressure, financial uncertainty, and limited access to campus services.

Most existing AI tools are English-first, bandwidth-heavy, and fragmented across many apps. We were inspired to create CampusSaathi AI as an India-first, multilingual companion that works even in low-connectivity environments and supports students across learning, careers, campus life, and mental well-being in one unified interface.

What it does

CampusSaathi AI acts as a single smart assistant for students:

  • Academic help: Notes summarization, exam planning, and doubt solving.
  • Career support: Resume feedback, internships, and scholarships.
  • Campus services: Hostel complaints, navigation, and event discovery.
  • Wellness: Mood check-ins, stress detection, and meditation guidance.
  • Accessibility: Multilingual chat, voice support, and low-bandwidth fallback modes.

How we built it

The system uses a modular AI pipeline:

  • Frontend: React / Flutter chat interface.
  • Backend: FastAPI + MongoDB.
  • Models: IndicBERT, IndicTrans2, and custom speech-to-text engines.
  • Routing Engine: Classifies user intent and language.

Mathematical Framework

We calculate the best response using the following scoring formula:

In this equation, (Sim(q,r)) represents the semantic similarity between the user query and the retrieved resource .

Backend Logic

def handle_query(text):
    lang = detect_language(text)
    intent = classify_intent(text)
    return generate_reply(text, lang, intent)

Challenges we ran into

  • Multilingual Accuracy: Supporting various Indian languages with high precision.
  • Optimization: Reducing model size for low data usage environments.
  • Ethics: Designing non-clinical, safe mental-health workflows.
  • Integration: Consolidating highly specific campus information sources.

Accomplishments

  • Working multilingual prototype.
  • Low-bandwidth text-only mode.
  • Unified academic, campus, and career assistant.
  • Mental-wellness module with integrated safety prompts.

What's next

  • Voice-only interface for feature phones.
  • WhatsApp & SMS deployment for wider reach.
  • Offline inference to allow on-device processing without internet.
  • Personalized Analytics dashboards for student learning patterns.

Built With

Share this project:

Updates