ROVIK / WORK / 02
FROM QUESTION BANK TO LEARNING INTELLIGENCE
IDRAAK is an AI-powered education platform designed to connect learning, assessment, performance intelligence and personalised preparation for competitive examinations.
idraak.pk
A student should not only know what they got wrong. The system should help determine what they should do next.
Executive overview
What was wrong, what ROVIK did, and what changed — then implementation, validation, and technology in the sections below.
Interactive education products need reliable access control, commerce, and content delivery without trusting the client for entitlements.
IDRAAK combines interactive learning experiences with server-authoritative checkout, Stripe verification, and database-driven entitlements.
A digital education product architecture where commerce, entitlements, and interactive learning are connected through verified server-side control.
SYSTEM REQUIREMENTS
Traditional exam preparation often separates study, MCQs, scores, incorrect answers and revision decisions. The student may receive a score but still have to manually determine next steps.
PRODUCT VISION
IDRAAK treats preparation as a continuous journey rather than independent pages — connecting learning activity, student data and intelligence.
STUDENT JOURNEY
The documented student journey spans discovery through improvement as a connected preparation cycle.
Establish examination context.
Subjects → Chapters → Topics → Subtopics
Answer questions and assessment items.
Examine sessions and mistakes.
Identify performance patterns.
Turn performance into preparation.
Use Exam Hall and other assessment flows.
Feed performance back into preparation.
LEARNING ARCHITECTURE
Structured hierarchy allows performance to be interpreted below the broad subject level.
This structure is designed so weaknesses can be identified with more context than a single score.
Performance signals
PERFORMANCE INTELLIGENCE
Documented intelligence-oriented product areas — presented as architecture, not performance results.
Performance data should become actionable information.
The current Learning DNA implementation is rule-based rather than itself being an LLM-generated profile.
IDRAAK AI
The AI Tutor is designed as a contextual learning assistant rather than a generic chatbot.
CONTEXT-AWARE AI
Server-side context is assembled from documented learning, performance and preparation signals before a tutor response is generated.
Context assembly follows documented server-side patterns such as buildTutorContext where applicable.
The AI architecture deliberately restricts what question information can be supplied to the model. A valid sessionQuestionId is required when contextualising an MCQ rather than accepting an arbitrary questionId.
Conversations are session-only in the current implementation — they are not currently persisted as permanent student chat history.
ASSESSMENT
Assessment experiences serve different roles across practice, diagnostics, simulation and review.
Knowing the material and performing under examination conditions are related but different problems.
IDRAAK includes a dedicated exam-oriented experience rather than treating preparation purely as knowledge acquisition.
STUDY PLANNING
Performance intelligence is designed to feed preparation recommendations and study planning.
Study planning is accessible via /study-plan within the application.
SUBSCRIPTIONS
Runtime entitlements are based on active student subscriptions and subscription_plan_capabilities. The database is the runtime authority for access.
basic
Core learning
test
Practice/testing capabilities
premium
Expanded assessment capabilities
complete
Full capability set including AI Tutor
Plan labels map to capability sets — access is enforced from subscription state, not UI visibility alone.
Protected functionality uses server-side guards. The AI Tutor API independently verifies entitlement before an AI request is made.
PAYMENTS
The public catalogue uses whole PKR values (e.g. PKR 4,999). Stripe expects amounts in minor units. The implementation converts when creating Stripe Checkout line items and normalises Stripe's returned amount when verifying payment. A mismatch was identified and fixed; the production build subsequently passed.
DATABASE & AUTH
CONTROL CENTRE
IDRAAK is not only a student-facing application — operational tooling supports subscriptions, assessments, intelligence and configuration.
Intelligence areas
ENGINEERING
Engineering work focused on connecting learning, intelligence, entitlements, payments and AI context into one coherent product.
Question → Topic → Performance → Mastery → Weakness → Preparation
Access must be enforced beyond frontend UI.
Server-authoritative checkout, catalogue validation, Stripe verification, webhook verification and idempotent fulfillment.
How can an AI tutor understand where the student currently is in preparation?
Learning, assessments, intelligence, subscriptions, payments, AI and operations became interconnected.
ROVIK / CONTRIBUTION
ROVIK contributed to the product's design and engineering across the application architecture, learning experience, student intelligence, subscription system, payment infrastructure and AI-assisted learning layer.
TECHNOLOGY
Next.js 16.3 · React 19 · TypeScript · PostgreSQL / Supabase · Supabase Auth · Next.js App Router · Stripe · Vercel · OpenAI provider architecture · npm
The documented production build contains approximately 176 API route handlers.
CURRENT STATE
Documented current application areas span learning, assessment, intelligence, subscriptions and operations.
The documented production build has been successfully compiled and deployed to idraak.pk.
Don't assume a feature works because the UI looks correct. Verify the underlying state and workflow.
The UI once displayed a stale payment-pending state even though membership was already active. The fix corrected the underlying state relationship rather than simply hiding the message.
DELIVERY & EVIDENCE
IDRAAK began with the familiar problem of examination preparation: students need content, practice questions and tests. The engineering challenge was connecting those pieces. The resulting architecture allows learning activity to feed performance intelligence, performance intelligence to inform preparation, and preparation to become context for further learning and AI assistance.
What did I score?
What does my performance tell me, what should I work on next, and how can the platform help me get there?
ROVIK / MORE WORK