Skip to content

AI-powered guidance platform helping Indian college students from Tier-2/3 colleges discover scholarships & opportunities through Hindi/Hinglish conversation. Built for AI For Bharat Hackathon - Track 6.

Notifications You must be signed in to change notification settings

AshharAhmadKhan/campusCOMPASS

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 

Repository files navigation

CampusCompass

AI-Powered Educational Opportunity Navigator for Indian Students

Hackathon Track Status

Quick Overview

CampusCompass is an AI-powered guidance platform that helps Indian college students from Tier-2/3 colleges discover and validate their eligibility for scholarships, exams, courses, and schemes—all through natural conversation in Hindi/Hinglish.

The Problem We're Solving

Millions of students from smaller colleges miss life-changing opportunities because:

  • They don't know these opportunities exist
  • Information is scattered across hundreds of websites
  • Eligibility criteria are complex and confusing
  • Most resources are English-only
  • No personalized guidance or mentorship available

Our Solution

An AI chatbot that:

  • Understands Hindi/Hinglish conversation (no language barriers)
  • Extracts student profiles naturally (no complex forms)
  • Validates eligibility accurately (hybrid rules + AI reasoning)
  • Provides 2-3 focused recommendations (prevents overwhelm)
  • Works on 2G networks (accessible on basic smartphones)

Challenge Track

Student Track 6: AI for Communities, Access & Public Impact

CampusCompass directly addresses educational equity by democratizing access to opportunities for underserved student communities.

Key Features

1. Multilingual Natural Language Processing

  • Processes Hindi, Hinglish, and English input
  • Handles regional language variations
  • Asks clarifying questions in user's preferred language

2. Intelligent Profile Extraction

  • Conversational data gathering (no forms!)
  • Extracts academic details, location, and preferences
  • Validates and normalizes information automatically

3. Hybrid Eligibility Reasoning

  • Rule-based filtering for objective criteria (accuracy)
  • LLM reasoning for complex/ambiguous requirements (flexibility)
  • Plain-language explanations for every decision

4. Actionable Recommendations

  • Maximum 3 opportunities per query (focused guidance)
  • Step-by-step application instructions
  • Deadline-aware prioritization
  • Verified links and contact information

5. Accessibility-First Design

  • Text-only interface (works on 2G)
  • Low bandwidth usage
  • Supports basic smartphones
  • Screen reader compatible

Architecture

┌─────────────────────────────────────────────────┐
│           User Interface Layer                  │
│         (Text-based Chat Interface)             │
└──────────────────┬──────────────────────────────┘
                   │ Hindi/Hinglish Input
                   ▼
┌─────────────────────────────────────────────────┐
│           Processing Layer                      │
│  ┌─────────────────────────────────────────┐   │
│  │   Conversation Handler (LLM-powered)    │   │
│  │   • Natural language understanding      │   │
│  │   • Profile extraction                  │   │
│  └──────────────────┬──────────────────────┘   │
│                     ▼                           │
│  ┌─────────────────────────────────────────┐   │
│  │  Eligibility Reasoner (Hybrid)          │   │
│  │  • Rule-based filtering                 │   │
│  │  • LLM validation for complex cases     │   │
│  └──────────────────┬──────────────────────┘   │
│                     ▼                           │
│  ┌─────────────────────────────────────────┐   │
│  │  Action Translator (LLM-powered)        │   │
│  │  • Recommendation ranking               │   │
│  │  • Next step generation                 │   │
│  └─────────────────────────────────────────┘   │
└─────────────────────────────────────────────────┘
                     │
                     ▼
┌─────────────────────────────────────────────────┐
│              Data Layer                         │
│  • Knowledge Base (JSON - opportunities)        │
│  • Session Data (SQLite - user context)         │
└─────────────────────────────────────────────────┘

Why Hybrid Architecture?

  • Rules ensure accuracy: No hallucination risk for eligibility decisions
  • LLM enables flexibility: Handles language understanding and complex reasoning
  • Cost-effective: Rule-based pre-filtering reduces expensive LLM calls by 60%

Cost & Scalability

User Load Cost per User Total/Month
1,000 (Demo) ₹1.08 ₹1,080
10,000 (Pilot) ₹0.42 ₹4,200
100,000 (District) ₹0.21 ₹21,000
1,000,000 (State) ₹0.17 ₹1,70,000

At 1M users/month: Just ₹0.17 per user — financially sustainable for government/NGO deployment at Bharat scale.

Cost Optimization Strategies

  • Intelligent caching of common queries
  • Rule-first processing before LLM calls
  • Batching multiple checks in single API calls
  • Model right-sizing (Claude-instant vs Claude-2)

Success Metrics

We measure impact through three concrete outcomes:

  1. Clarity Metric: 80% of users identify at least one actionable opportunity
  2. Action Metric: 50% of users click through to opportunities within 7 days
  3. Trust Metric: Users return with follow-up questions within 7 days

These metrics are:

  • Testable during demos
  • easurable in pilot phase
  • ✅ Aligned with real-world impact

Tech Stack

  • Backend: FastAPI (Python)
  • AI/LLM: Claude API (Anthropic)
  • Database: SQLite (sessions), JSON (knowledge base)
  • NLP: Multilingual processing (Hindi/Hinglish/English)
  • Deployment: Lightweight, cloud-agnostic design

Repository Structure

campusCOMPASS/
├── requirements.md    # Detailed functional requirements
├── design.md         # System architecture and design decisions
├── tasks.md          # Project tasks and milestones
└── README.md         # This file

Target Audience

Primary Users

  • Students from Tier-2/3 colleges
  • Semi-urban and rural backgrounds
  • First-generation learners
  • Limited English proficiency
  • Basic smartphone users with 2G/3G connectivity

Geographic Focus

  • All states of India
  • Priority: Uttar Pradesh, Bihar, Madhya Pradesh, Rajasthan (high student populations in smaller colleges)

Impact Potential

Scale of the Problem

  • 40+ million college students in India
  • 70%+ from Tier-2/3 colleges
  • Thousands of scholarships/opportunities available annually
  • <10% awareness among target students

Our Impact Vision

If CampusCompass helps just 1% of Tier-2/3 students access one opportunity they would have missed:

  • 280,000+ students annually benefit
  • ₹100+ crore in scholarships/opportunities unlocked
  • Multiplier effect through peer sharing

Privacy & Ethics

  • Minimal data collection: Only essential academic information
  • No PII storage: Avoids storing sensitive personal identifiers
  • User control: Easy session data deletion
  • Transparent AI: Clear explanations for all recommendations
  • Bias awareness: Regular audits for regional/demographic fairness

Current Status

Hackathon Submission Stage

  • Requirements documented
  • System design completed
  • Architecture validated
  • Prototype development in progress

Team

ThreadFall - Student Hackathon Team

  • Focus: Clean code, solid system design, rapid prototyping
  • Philosophy: Solving real problems with practical AI

Contact


License

This project is developed as part of the AI For Bharat hackathon.


Acknowledgments

Built for Track 6: AI for Communities, Access & Public Impact because every student deserves a compass to navigate their future.

Powered by: Claude AI (Anthropic) | Hosted by: AI For Bharat Hackathon | Built with: ❤️ for Bharat's students


"Education is the most powerful weapon which you can use to change the world." - Nelson Mandela


Hackathon Links

About

AI-powered guidance platform helping Indian college students from Tier-2/3 colleges discover scholarships & opportunities through Hindi/Hinglish conversation. Built for AI For Bharat Hackathon - Track 6.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published