AIDE for Grades 9–12

This is the optional Week 0 — set up your free AI accounts (ChatGPT first, then others you'll meet in later weeks), then take the skip-ahead Python diagnostic. If you already know Python, you can skip straight to Week 1. Week 0 does not count toward your certificate.

Every example uses something you already touch — exam marks, mandi prices, GST on a kirana bill, IRCTC fares, and real district-level rainfall data from data.gov.in. No `foo`/`bar` placeholder code anywhere in this week.

Learn how to talk to ChatGPT so it actually does what you want — anatomy of a prompt, role / context / task / format, and the iteration habit that separates good AI users from great ones.

Every example uses a Class 9 NCERT maths tutor scenario — translating questions into Hindi, showing step-by-step working — so the prompting techniques land in a context students already live in.

Master every capability of free ChatGPT — chat, voice, image generation, file upload, and web browsing — and learn the power-user moves that turn it into a full study companion. By the end of this week you will have transcribed a real audio clip, generated an image from a text prompt, and used an existing GPT from the store.

A Hindi-medium student uses ChatGPT's voice feature on a ₹8,000 Android phone to transcribe a class lecture in real time, then asks follow-up questions in Hindi without typing a single word.

Use the best free AI tools for each study job — grounded research with NotebookLM, concept explanation with Gemini or Claude, flashcards with NotebookLM or Anki, essay drafting and critique with Claude, and notes from recordings with Otter or Sarvam. Always verify AI answers. Build one real study aid from your own syllabus by the end of the week.

A Class 10 Hindi-medium student preparing for CBSE boards uses NotebookLM loaded with NCERT chapters to get grounded answers, then builds Anki flashcards from the same source — so every card is traceable back to the textbook.

Use free AI specialist tools to produce a real project asset — a slide deck, a polished report, or a mini-website — in one sitting. Learn which tool wins each job and why, then keep a durable backup move for every task.

We build a science-fair project asset for a Class 10 student — a slide deck presenting air-quality data from a local CPCB station, a one-page report for the teacher, and a landing page to share findings with the school.

Use the best free specialist AI tool for each real-world job — research with verified citations, data in Google Sheets, transcription to action items, and translation into Indian languages. Chain these tools together to complete a full task end-to-end.

We research CPCB air-quality data and Agmarknet mandi prices, verify every source before trusting it, translate findings into Hindi for a wider audience, and transcribe a class discussion into clean notes and action items.

Two halves in one week. First, crack open how a large language model actually works — tokens, next-token prediction, training data, hallucination, temperature, and context windows — so you stop treating AI as magic and start controlling it. Second, use Python in the browser to load a real Indian CSV, run a simple ML model, and read its accuracy including F1.

We load an IMD district-level rainfall CSV and an Agmarknet mandi-price CSV in the browser, explore them with pandas, then train a classifier and read its accuracy and F1 score on real Indian data.

Build a real machine-learning model end to end — from raw data to saved file. You will learn how a decision tree "learns" from features and labels, why a train/test split is the single most important idea in honest ML evaluation, and how to read accuracy without fooling yourself. Weekly output: a trained Class-10-pass classifier, an honest accuracy note, and a "what I built" paragraph.

Predict whether a student passes Class 10 from study hours, attendance percentage, and private tuition — a dataset rooted in the Indian school system.

AI can be biased, privacy-invasive, and weaponised for misinformation — and you need to know how to spot it. We'll analyse real Indian incidents (Aadhaar exclusion, deepfake scams, UPI fraud, board-exam rumours) to build a concrete framework for using AI with integrity, verifying its output, and knowing when to keep it out entirely.

From Aadhaar biometric exclusion of manual labourers to deepfake audio scams targeting Indian families and viral board-exam misinformation on WhatsApp — India's AI risks are real, recent, and worth understanding before you build.

Stop copying the same instructions into every chat. Learn what an AI "agent" actually means, build a reusable personal assistant the free way — a system-prompt template, ChatGPT Projects, or a Gemini Gem — and then chain LLM steps into a multi-step workflow that automates a repetitive task you face every week.

We build a reusable "Class 9 Maths Tutor" assistant — one that always responds step-by-step in Hindi or English — and then automate the weekly task of converting NCERT exercise answers into a revision flashcard set, all without spending a rupee.

Pull everything together into one project you can point to. Pick a real Indian problem, choose the right tools from anything you've learned — an LLM app, a free-tool workflow, a Python ML model, a study-aid site, an automation, a data deck — build a working v1, polish it, deploy it to a live link, and present it with a writeup and an ethics note. This is the week your portfolio becomes something you can talk about in an interview.

The capstone is your call. Solve a problem someone actually has — not a Kaggle leaderboard. Past student capstones include: a Marathi-English code-switched complaint router for a municipal corporation, a vegetable price predictor for a Pune mandi, a school-dropout risk score for a Bihar district, a Hindi-language nutrition assistant for ASHA workers, a Notion-based study planner for JEE students using GPT prompts, and a free Glitch site explaining a local government scheme in plain Hindi. Any approach, any tools — the test is: did you finish it, ship it live, and honestly describe what it gets wrong?