MULYAYON
AI-Powered Assessment & Academic Evaluation Platform
An AI-powered assessment platform designed to help educators evaluate assignments, review answers, provide rubric-calibrated feedback, and track student performance across academic institutions.
Product Vision
A comprehensive assessment intelligence platform engineered to modernize academic evaluation for Bangladesh's educational ecosystem. MULYAYON streamlines exam creation, automated rubric-aligned grading, teacher review workflows, and deep student performance analytics.
The Problem
Traditional academic assessment in large classroom settings is plagued by manual grading delays, inconsistent evaluation criteria, and a lack of granular actionable feedback for students and institutional leadership.
The Solution
Architected an end-to-end evaluation engine utilizing Next.js, Express, and PostgreSQL with an integrated AI evaluation pipeline. Teachers upload student scripts or assessments, receive rubric-calibrated AI draft grading with line-by-line feedback, and review/override decisions via an ergonomic grading dashboard.
How It Works
The end-to-end evaluation and review lifecycle engineered for academic staff.
Assessment & Rubric Ingestion
Faculty configure course rubrics, grading criteria, and acceptable score ranges tailored to curricular guidelines.
Script Processing & AI Evaluation
Student submissions are processed through the inference pipeline to generate provisional line-by-line marks and constructive feedback.
Educator Review & Calibration
Teachers review AI recommendations on an ergonomic grading console, with full override control and manual commentary adjustment.
Analytics & Performance Reporting
Aggregated performance trends, common error patterns, and individual mastery distributions are exported for institutional oversight.
Architecture Highlights
Technical patterns and boundary choices prioritizing throughput and integrity.
Modular micro-service communication between Next.js App Router frontend and Node/Express AI processing worker
PostgreSQL relational schema managed via Prisma with strict foreign key constraints for academic integrity
Optimistic UI updates for high-speed grading interactions with background synchronization
Strict runtime validation via Zod across all API endpoints and client submission forms
Key Features
AI-Assisted Assessment
Rubric-based script evaluation with instant feedback generation
Teacher Review Workflow
Side-by-side rubric verification with real-time score adjustment
Bangladesh-Focused Workflow
Structured for local curriculum formats and grading conventions
High-Density Analytics
Class-wide performance heatmaps, rubric mastery charts, and exportable reports
Granular Role Permissions
Multi-tenant access controls for teachers, moderators, and students
Responsive Grading Console
Keyboard-first interface engineered for rapid evaluation sessions
Engineering Trade-Offs
Conscious technical choices made to balance speed, isolation, and developer experience.
Next.js App Router for Front Office
Utilized React Server Components for near-zero client bundle overhead on data-heavy dashboards, paired with client boundaries for optimistic grading actions.
Decoupled Express AI Worker
Separated compute-heavy LLM evaluation queues from the primary web server to guarantee low-latency HTTP responses during batch grading sessions.
Prisma ORM with PostgreSQL
Enforced database-level referential integrity across submissions, rubrics, and grade modifications with full TypeScript type safety across queries.
Technologies Utilized
Production Deployment & Impact
Active production deployment running at mulyayon.vercel.app, modernizing script review cycles and providing instructors with rubric-aligned grading intelligence.