UPSC Prelims Test Analyzer (Web App)
Budget: ₹1,500 – ₹12,500 INR
UPSC Prelims Test Analyzer (Web App)
Overview
Build a web-based UPSC Prelims Test Analyzer that converts any coaching/self-made Prelims test PDF into:
Accurate question extraction (≈100 questions)
Question-by-question review with AI-generated solutions
Detailed analytics and leaderboard
Clean, modern gradient-based UI
This is not a test-taking or social platform. Scope is strictly limited to features below.
Tech Stack (Preferred / Existing)
Frontend: React + Tailwind CSS
Backend: FastAPI
Database: MongoDB
OCR: High-accuracy OCR pipeline (see below)
LLM: API-based (for solution generation)
Payments: Razorpay
Fonts: Inter / Space Grotesk (Google Fonts)
Core Features (Mandatory)
1. Authentication
Phone number login (OTP-based)
Protected routes for logged-in users
2. Test Upload
Upload question paper PDF (scanned or digital)
Optional answer key upload (PDF or manual entry)
No auto-fetching of answer keys
3. OCR & Question Extraction (Critical)
Extract ≥95 out of 100 questions reliably
Must extract:
Question number
Question text
Options A–D
Must handle:
Multi-page questions
Statement-based questions
Wrapped options
Maps/diagrams
OCR pipeline:
Try text extraction first
Fallback to OCR (300–400 DPI)
Layout-aware parsing
Auto-retry if question count < 90
Cache extracted questions to avoid reprocessing
4. Question Attempt Flow
For each question, user can mark:
Selected option (A/B/C/D / Unattempted)
Attempt type:
R1: Confident
R2: Elimination / 50–50
R3: Fluke / Blind
NA
5. AI-Generated Solutions (No Manual Upload)
System auto-generates solutions using LLM
400–500 words per question
Must include:
Correct option
Explanation
Brief elimination of wrong options
Solutions cached per question
No “solution upload” page
6. Question-by-Question Review Page
Each question must show:
Question text + options
User answer vs correct answer (color-coded)
Attempt round badge
Subject + difficulty tag
Full solution text
Copy button (copies question + options + correct answer + solution)
7. Scoring
+2 correct
−2/3 incorrect
0 unattempted
Net score used everywhere
8. Analytics Dashboard
Overall score, accuracy, net marks
Subject-wise performance
Decision-quality metrics:
Confidence accuracy
Elimination efficiency
Fluke dependency
Negative mark vulnerability
Remove any “Strategy” tab
9. Leaderboard
Leaderboard fields:
Name
Test name
Test code
Coaching institute
Marks (net)
Rules:
Same test code = same cohort
Rank by marks
Tie-break by accuracy
Opt-in, anonymous allowed
10. Multi-Test Tracking
Track performance across multiple tests
Show trends in score, accuracy, subjects
11. Payments (Razorpay)
Integrate Razorpay for paid access
One-time or subscription-based access
Payment-gated features:
Full analytics
Full solutions
Leaderboard visibility
UI / Design Requirements
Gradient mesh backgrounds
Glass-morphism cards (backdrop blur)
Indigo → purple gradient buttons
Inter (body) + Space Grotesk (headings)
Modern, clean, academic aesthetic
Responsive (mobile + desktop)
Explicitly Out of Scope
Study buddy / social features
Chat / messaging
Practice questions / test creation
Study plans / recommendations
Auto-fetching answer keys
Gamification
Deliverables
Fully working web app
Clean, readable code
OCR pipeline with retries
LLM integration for solutions
Razorpay integration
Basic documentation for setup
Success Criteria
Consistently extracts ~100 questions from real coaching PDFs
Solutions are generated automatically (400–500 words)
Smooth user flow from upload → analysis → leaderboard
UI matches modern gradient/glass design
Overview
Build a web-based UPSC Prelims Test Analyzer that converts any coaching/self-made Prelims test PDF into:
Accurate question extraction (≈100 questions)
Question-by-question review with AI-generated solutions
Detailed analytics and leaderboard
Clean, modern gradient-based UI
This is not a test-taking or social platform. Scope is strictly limited to features below.
Tech Stack (Preferred / Existing)
Frontend: React + Tailwind CSS
Backend: FastAPI
Database: MongoDB
OCR: High-accuracy OCR pipeline (see below)
LLM: API-based (for solution generation)
Payments: Razorpay
Fonts: Inter / Space Grotesk (Google Fonts)
Core Features (Mandatory)
1. Authentication
Phone number login (OTP-based)
Protected routes for logged-in users
2. Test Upload
Upload question paper PDF (scanned or digital)
Optional answer key upload (PDF or manual entry)
No auto-fetching of answer keys
3. OCR & Question Extraction (Critical)
Extract ≥95 out of 100 questions reliably
Must extract:
Question number
Question text
Options A–D
Must handle:
Multi-page questions
Statement-based questions
Wrapped options
Maps/diagrams
OCR pipeline:
Try text extraction first
Fallback to OCR (300–400 DPI)
Layout-aware parsing
Auto-retry if question count < 90
Cache extracted questions to avoid reprocessing
4. Question Attempt Flow
For each question, user can mark:
Selected option (A/B/C/D / Unattempted)
Attempt type:
R1: Confident
R2: Elimination / 50–50
R3: Fluke / Blind
NA
5. AI-Generated Solutions (No Manual Upload)
System auto-generates solutions using LLM
400–500 words per question
Must include:
Correct option
Explanation
Brief elimination of wrong options
Solutions cached per question
No “solution upload” page
6. Question-by-Question Review Page
Each question must show:
Question text + options
User answer vs correct answer (color-coded)
Attempt round badge
Subject + difficulty tag
Full solution text
Copy button (copies question + options + correct answer + solution)
7. Scoring
+2 correct
−2/3 incorrect
0 unattempted
Net score used everywhere
8. Analytics Dashboard
Overall score, accuracy, net marks
Subject-wise performance
Decision-quality metrics:
Confidence accuracy
Elimination efficiency
Fluke dependency
Negative mark vulnerability
Remove any “Strategy” tab
9. Leaderboard
Leaderboard fields:
Name
Test name
Test code
Coaching institute
Marks (net)
Rules:
Same test code = same cohort
Rank by marks
Tie-break by accuracy
Opt-in, anonymous allowed
10. Multi-Test Tracking
Track performance across multiple tests
Show trends in score, accuracy, subjects
11. Payments (Razorpay)
Integrate Razorpay for paid access
One-time or subscription-based access
Payment-gated features:
Full analytics
Full solutions
Leaderboard visibility
UI / Design Requirements
Gradient mesh backgrounds
Glass-morphism cards (backdrop blur)
Indigo → purple gradient buttons
Inter (body) + Space Grotesk (headings)
Modern, clean, academic aesthetic
Responsive (mobile + desktop)
Explicitly Out of Scope
Study buddy / social features
Chat / messaging
Practice questions / test creation
Study plans / recommendations
Auto-fetching answer keys
Gamification
Deliverables
Fully working web app
Clean, readable code
OCR pipeline with retries
LLM integration for solutions
Razorpay integration
Basic documentation for setup
Success Criteria
Consistently extracts ~100 questions from real coaching PDFs
Solutions are generated automatically (400–500 words)
Smooth user flow from upload → analysis → leaderboard
UI matches modern gradient/glass design