> INITIALIZING KITE AI CORE...
Hero Background
Kerala Infrastructure and Technology for Education

The KITE Model for
Responsible AI

Insights from Kerala's globally recognized indigenous ecosystem for integrating AI into schools — setting the benchmark for ethical technology adoption, presented to the Government of India.

2.5 Decades of ICT@Education

A Holistic Single Ecosystem

KITE integrates five interlocking pillars into a single eco-system — a uniquely Indian, public-sector approach that has matured over 25 years.

Capacity Building

Teachers themselves became the trainers — no external IT vendors. Trainings span ICT, hardware and content development for teachers, students, parents and the public.

Content Development

Adoption of FOSS in education, IT made a compulsory subject, ICT textbooks for all classes, and the Samagra Resource Portal.

Hardware Deployment

Deployment of ICT hardware, broadband connectivity, call-centre support, and an eWaste disposal program.

VICTERS

VICTERS & Support

VICTERS educational channel, Little KITEs IT clubs, School Wiki, eCube Language Lab and the Sahitham Portal.

eGovernance

End-to-end e-Governance for schools — bringing transparency, accountability and a coordinated state-wide platform.

Glimpses in Action

Transformation on the Ground

AI Training for Parents by LK Members
Student Activity
AI Training for Parents by LK Members
AI Training for Parents by LK Members
KITE Program
AI Training for Parents by LK Members
AI Training for Parents by LK Members
Classroom Action
AI Training for Parents by LK Members
Robotics Action
AI Training for Parents by LK Members
School Activities
AI Training for Parents by LK Members
Status Report

AI Learning Stages & Beneficiaries

How AI learning is being implemented by KITE — five sequenced stages, each with a defined beneficiary group.

1

AI Training for Teachers

Awareness on AI tech, teaching tools, pedagogical approaches & Responsible AI.

Teachers
Completed
2

AI in School Curriculum

Introduction from Class 7 — practicals, robotics, ethics & inclusive learning.

Teachers & Students
Completed
3
Samagra Plus

Samagra AI Engine

Integration of LLMs to provide curriculum-aligned services to students and teachers.

Teachers
In Progress
4

AI Insights for Public

Video series providing AI literacy via the KOOL platform & KITE VICTERS channel.

General Public
Completed
5

AI-enabled Research

Tools and practices to be integrated into research workflows across institutions.

Scholars & Faculty
In Pipeline
Historical Journey

Little KITEs Milestones

2025

Artificial Intelligence (AI) training for Little KITEs members.

Advanced Robotic Kits (Moving Robots) to be deployed state-wide.

2024

UNICEF study terms Little KITEs a Global model in Education.

29,000 Robotic Kits deployed. Robotic Fests conducted in 2,000 schools.

2023

State-wide study on Little KITEs outcomes conducted by UNICEF.

2022

Little KITEs train 4 Lakh Mothers in Cyber Safety.

Sub-district camps for 15,000 LK student members conducted.

2020

5% Grace marks for Class 10 Little KITEs members.

District camps for 1,150 LK members conducted.

2019

Camera Training for Little KITEs members. IoT & 3D modeling workshops.

Digital magazines of 2,000 schools made by LK members published.

2018

Hi-school Kuttikoottam renamed to Little KITEs IT Clubs.

Launched by Hon. CM. Free Software Install Fests state-wide.

2011

ANTS (Animation Training) for 12,677 students. State-wide hardware training.

2010

SSITC (School Students IT network) formally formed.

UNICEF

National Record

Largest Pupil IT Club in India with 2 lakh+ active mentors.

Global Model

Featured by UNICEF as a global benchmark for public ed-tech.

Infrastructure

2.5 Decades of Legacy

2.5 Lakh+ Laptops

70,000 Projectors across 16,000 schools state-wide. High-speed broadband in all classrooms.

2 Lakh+ Teachers

Teachers trained in ICT, creating a self-reliant mentoring ecosystem state-wide.

VICTERS

VICTERS Hub

India's first educational channel reaching every home 24/7 with educational content.

Why Responsible AI Matters

Global Voices on AI & Children

"AI's rapid evolution makes it difficult for policymakers to keep up, but it is necessary for advocates for children to take stock of how well these technological advances benefit all children, especially the poorest and hardest to reach."

— UNICEF Innocenti

"Because of the data on which generative AI is trained, children and adults experience similar or worse bias and discrimination using AI tools as they do in their offline lives."

— UNICEF on Generative AI & Children's Rights

A 2022 Human Rights Watch investigation found that several EdTech platforms used in India shared sensitive data with third-party advertisers, exposing children to profiling & long-term surveillance.

— Human Rights Watch, 2022

UNESCO Framework

5 Global Challenges in AI for Education

01. Policy

Cohesive policies across ecosystem.

02. Inclusion

Fair access to AI tools for every learner.

03. Training

Integrating AI effectively into pedagogy.

04. Data

Avoiding biases & reliable insights.

05. Ethics

Privacy, accountability and fairness.

The KITE Mandate

Eight Challenges Kerala Set Out to Solve

As featured in The Hindu — "How Kerala is setting a global benchmark for ethical AI in education."

Curriculum Gap

Publicly authored, age-appropriate material.

Market Bias

Avoiding vendor-lock by building on FOSS.

Data Security

Local servers & zero third-party sharing.

Digital Divide

Offline-ready tooling for every classroom.

Value Integration

Embedding ethics & inclusivity.

Unpredictable AI

RAG architecture for zero hallucinations.

Pedagogy Erosion

AI assists, never replaces, the human mentor.

Child Data

No profiling, no advertiser data.

AI in School Curriculum

How AI is Gelled into Classes 7–10

AI is not a side activity — it is integrated into the official ICT textbook, with a clear progression from foundational concepts in Class 7 through robotics in Class 10.

How AI Works

  • Introduction to how AI technology works.
  • Computer Vision & Face Detection.
  • Children build a program that recognizes 7 facial expressions using a pre-trained model.
  • Block programming via Pictoblox.
  • Leveraging gesture-based software for CWSN.

Machines that Learn

  • Supervised, Unsupervised & Reinforcement Learning.
  • Activity: build an ML model to recognize Indian Sign Language (ISL).
  • Develop an interactive ISL letters game.
  • Hands-on: Dataset preparation → Training → Testing.

Intelligent Machines

  • Natural Language Processing (NLP) & how it works.
  • Large Language Models (LLMs) — interacting with an LLM.
  • Prompt Engineering, text generation, image creation with Gen-AI.
  • Responsive AI & AI Agents in learning.

The World of Robots

  • Fundamentals of Robotics & AI.
  • Build a model of a smart door that opens via face recognition.
  • Control electronic devices with AI & ML.
  • Pre-trained model + Arduino + webcam + servo motor.
Public AI Literacy

KOOL — KITE's Open Online Learning

A MOOC-based training platform developed by KITE, now offering a 4-week 'AI Essentials' course to the general public — covering ethical use, fake news and everyday applications.

1,216

Individuals from all walks of life trained so far in 3 batches.

4 weeks

'AI Essentials' course on ethical use, deepfakes & daily AI applications.

1 : 20

Dedicated mentor for every 20 participants — true human-led learning.

100%

FOSS-based content, authored entirely by Kerala teachers.

What teachers actually learned in AI

Summarization & Media

Image generation, Text-to-Speech and Speech-to-Text for classroom use.

Prompt Engineering

Prompting techniques for accurate text and image generation.

Question & Resource Gen

Preparing questions, tables, crosswords and worksheets at scale.

Classroom Strategy

Explainable AI, deepfake awareness & bias detection in algorithms.

Samagra Plus

Samagra Plus AI — Kerala's Indigenous AI Engine

A RAG-based AI system that responds using our own knowledge base aligned with curriculum — retrieving curriculum-specific content from a vetted database before responding, ensuring relevance, accuracy and zero hallucinations.

100% FOSS Tech Stack

Built with LangChain, ChromaDB, FastAPI backend, Streamlit interface and OpenAI Whisper for speech. Zero data sharing — entirely on private datasets.

Curriculum-Aligned RAG

Describes learning concepts with images, videos and interactives drawn from the Samagra Plus content portal — accurate, age-appropriate, vetted.

Adaptive Feedback

Identifies weak areas based on student inputs to interactive quizzes and produces personalised feedback for both students and teachers.

Samagra Plus

Six AI Services Built for Indian Classrooms

NEW

Teaching Manual (TM) Generator

Auto-generates Teaching Manuals mapped to curriculum, Learning Outcomes and digital resources in the Samagra Plus portal. Suggests classroom activities and saves teacher preparation time.

BHASHINI

Multilingual Translation

Translates Teaching Manuals into English, Malayalam, Tamil and Kannada, powered by BHASHINI APIs — ensuring accessibility across diverse learners.

24 / 7

AI Chatbot — Concept Explanation

A curriculum-aligned chatbot that explains topics using approved digital resources and age-appropriate content — available round the clock for student support.

QUIZ

Quiz Generator

Generates interactive quizzes via LLMs based on curriculum content. Supports multiple difficulty levels and instant feedback for formative assessment.

HOTS

Higher Order Thinking Skills

Generates descriptive HOTS questions that encourage analysis, reasoning and creativity. AI-based answer evaluation via the EAI Engine.

VOICE

Let's Talk — Speaking Practice

Interactive text and voice AI bots for speaking practice on common topics: home & family, pets, travel, daily life — building communication and confidence.

Higher Secondary

Key to Entrance — AI for Competitive Exams

10,000+ AI Questions

Designed for higher-secondary students with over ten thousand AI-generated entrance questions across multiple subjects.

Personalized Study Plan

AI-driven personalised learning paths with adaptive recommendations based on student's performance analytics.

Building India's Own AI

Engineering Challenges & How KITE Overcame Them

Kerala chose to develop its own AI engine inside the public sector — here are the four big walls we hit, and how we walked through them.

Challenge: Need for costly high-end GPUs to run modern LLMs.

Solution: Optimised FOSS stack & selective model use; sharing infrastructure across services to amortise GPU cost.

Challenge: GPU-cloud providers unavailable in the government sector — dependency on private clouds like AWS.

Solution: Architected for portability across providers; pursuing on-prem and indigenous public-sector GPU options.

Challenge: LLMs have known limitations in solving Mathematics & Physics problems.

Solution: Availed the DeepSeek API for advanced reasoning while keeping core flow on the Kerala stack.

Challenge: Dataset creation is time-consuming — cannot rely on textbooks alone; need to consolidate textbooks, teacher books and authorised resources, and map every Samagra interactive.

Solution: Teacher-authored, curated, vetted RAG knowledge base — purpose-built for Kerala's curriculum.

Responsibility

Safe & Private AI Environment

Protecting Rights

Answering HRW 2022 surveillance concerns through vetted datasets and local server processing — no profiling, no third-party advertisers.

Offline Access

Deploying powerful local AI models for schools without robust internet connectivity — a current focus area for KITE.

Cost of External APIs

Reducing recurring DeepSeek API costs by training and hosting our own reasoning models within the Kerala public-sector stack.

A Public-Sector, Indigenous, Ethical AI Model

From the first ICT textbook to Samagra Plus AI — Kerala's KITE model demonstrates that responsible, world-class AI in education can be built inside the public system, in Indian languages, on FOSS, for every child.

Download Full PDF Presentation