Apar AI LMS
A school LMS that arrivesfull, not empty.
How Aparsoft built Apar AI LMS: three dashboards, for the school office, the teacher and the student, on a library of 3 lakh+ learning assets made from NCERT books. The library is written in advance with AI and is ready on day one; AI works live where a student or a teacher asks for help.
Watch it work
Lesson plans, assessments and multilingual learning for NCERT-aligned schools.
The challenge
Most school software arrives empty.
Folders to fill
Months of content work before the first class can use it.
Tests to type by hand
Every test and paper made from scratch.
Copies to check, late at night
Marking eats the teacher's evening.
Weak topics found too late
After the exam, not before it.
The idea
Technology in education, not education on technology.
How it works
One school, three dashboards, one chapter
Real screens from the demo school, with sample data.
The school office sees the whole school
Students, teachers, classes and who signed in today, on one dashboard.
- Create classes and give them their teachers
- Invite teachers and students
- Your school's name and logo, on every screen
Take one chapter: acids, bases and salts
The teacher unlocks it when it is taught. Notes, lesson plans and concept maps are already inside.
- Lesson plans, summaries and concept maps
- Keywords, notes and textbook questions
- Shared with a class in one step
Six lesson plans for every chapter
Blended, flipped, project-based, collaborative, inquiry-based and direct instruction.
- Each with its own activities and timing
- Made from the chapter's own text
- Marked as AI-generated, for the teacher to use or adapt
Concept maps that show how ideas connect
The acid–base–salt cycle, as a diagram with its key concepts.
- The diagram and its key concepts, side by side
- Labelled by level of detail
- Ready for the class on day one
A test takes three clicks
Select, add, distribute.
- Questions from the chapter's ready bank
- MCQs are graded automatically
- Exam papers come with quality scores, for the teacher to approve
- 1Select the questions
- 2Add them to the test
- 3Distribute it to the class
- 4MCQs are graded automatically
- 5The teacher reviews the results
A drawing of the flow, not a screen.
Weak topics show up before the exam
Weak topics show up in class analytics. Before the exam, not after it.
- Students needing attention, flagged
- Each subject's scores, class by class
- Computed from the students' real work
Students open the same chapter
Their subjects, their progress and what is due next.
- Shared notes appear at once
- AI Explain turns a hard idea into simple steps
- Practice with AI feedback, and an AI viva with a virtual teacher
The whole school's numbers, live
Computed live from submissions, completions and viva sessions.
- Average score and participation
- How the class is spread, from excellent to needs improvement
- By class, subject, teacher and student
How we engineered it
Six ideas behind a library that arrives full
How a chapter becomes ready classroom material, and where people stay in charge.
Everything starts from the chapter
Notes, lesson plans, summaries, concept maps and questions are all made from the chapter's own text, not from an open prompt. That keeps every asset true to what the book teaches.
Made in steps, checked, then fixed
Each kind of material is made in stages: understand the chapter, write, check it against that kind's own quality bar, and fix what fails. A lesson plan and a concept map are held to different bars.
Ready before the school arrives
The library is built in advance, so a school starts with lakhs of assets on day one. AI works live only where it helps a person: explaining, practising, a viva, checking answers.
Assessment is a system, not a chat box
Questions, answer checking, quiz analysis and the viva each have their own process, with checks and a next step built in. The product responds to a student's work, not to a clever prompt.
People decide where trust matters
Content that affects trust can pause for a teacher's review before it is saved; low-risk steps run on their own, with checks. A teacher approves a paper before it reaches the class.
One set of numbers for everyone
Every percentage is computed live from real work and normalised per assessment, so the school office, the teacher and the student see the same picture, each from their own side.
What the build proves
A school that starts full, and grows from there
- A school starts with its content ready, not with empty folders
- MCQs are graded automatically; papers wait for the teacher's approval
- Weak topics show up in class analytics before the exam
- Nothing to install: it updates itself
More screens
Every screen, from the demo school
Class 10 Science, Acids, Bases and Salts: concept maps, lesson plans and summaries, ready in the chapter. Demo school, sample data.
How it's built
The stack, and how we use it
Proven open-source tools, used the way they are meant to be used.
Next.js 16 · React 19
Three dashboards on the App Router, with TanStack Query caching and refreshing every call to the server.
Django REST Framework
One documented API for the office, the teacher and the student, its OpenAPI schema generated from the code.
PostgreSQL + pgvector
School data and semantic search in one database, so there is no second store to keep in sync.
LangGraph + LangChain
Content and grading as stateful workflows with checks and retries, and any model behind one interface.
Django Channels
AI Explain, answer checking and the viva stream back live over WebSockets, word by word.
Celery + Redis
Generation, grading and reports run in background queues; scheduled jobs run on Celery Beat.
Bhashini
The Government of India's language platform, for translation and speech in Indian languages.
Docker + GitHub Actions
Containerised services, built and deployed by CI, with media in S3-compatible object storage.
