The Class 10 AI Project
Part C of Artificial Intelligence is the practical half of the subject — 50 of its 100 marks. Fifteen of them come from one piece of work you choose: a project, a field visit, or a portfolio. This page explains what is asked, and gives fifteen worked AI projects to learn from.
The projects on these pages are sample projects, written for reference and self-learning. Each one is a practical illustration of the AI Project Cycle — how a real problem is scoped, how data is collected and explored, and how a model is built, tested and put to use.
They are not meant to be submitted as they are. Most of the datasets are samples made by a program, not real records. Use these projects to understand the method, then build your own: choose a problem you care about, collect your own data, and write the report in your own words. That is what the project marks — and the viva — are really testing.
1Where the marks are
The 2026-27 curriculum splits Part C like this:
| Part C | Marks |
|---|---|
| Practical file with at least 15 programs | 15 |
| Practical examination — Statistical Data, Computer Vision, NLP, Advance Python | 15 |
| Viva voce | 5 |
| Project work / field visit / student portfolio (any one)this page | 10 |
| Viva voce on the projectthis page | 5 |
| Total | 50 |
So the project is worth 15 marks: 10 for the work and 5 for the viva about it. You only do one of the three choices. A field visit means taking part in something like an AI for Youth bootcamp or an AI exhibition. A portfolio continues from Class 9 and needs at least five AI activities. This page is about the third choice, the project.
2What the curriculum asks for
Problem scoping, data acquisition, data exploration, modelling, evaluation and deployment — all six stages.
The curriculum says it in so many words: relate the project to the Sustainable Development Goals.
Statistical Data, Computer Vision or Natural Language Processing. It gives one sample project for each.
The curriculum's three sample projects are built in no-code tools — Orange Data Mining and Teachable Machine. Every project here is done both ways: in Python, with the pandas, NumPy and matplotlib of Unit 7, and step by step in the no-code tool. The Python version is what lets every number be checked; the no-code version is what the curriculum's practicals use.
3Using these projects honestly
These are worked examples, not projects to hand in. Copy one, and the viva will show it in a minute: you will be asked why you chose your features, what your confusion matrix means, and what you would change. Use them to see what a finished project looks like, then build your own — a different problem, or the same problem on data you collected.
- Say where your data came from. Most datasets here are LambdaLab samples made by a program, and every page says so. One is a real public dataset, credited in full. Yours must be credited too.
- No names in your data. Use roll numbers or sample numbers. Personal data needs care — that is part of the ethics the curriculum teaches.
- The PDF has blank lines for the names on its cover, certificate and acknowledgement. Fill in your own.
4The projects
Fifteen projects, grouped by domain — seven, four and four, roughly in proportion to the lab hours the curriculum gives each: 28 hours for Statistical Data, 20 for Computer Vision and 7 for NLP. The curriculum's own sample project comes first in each group. Between them they draw 72 charts and carry 130 test cases, and every chart, number and test result on these pages comes from actually running the code.
Statistical Data
The data is numbers in a table — a CSV file. Curriculum sample: Prediction of Palmer Penguin species.
Four measurements, three species, and a model that tells them apart — the curriculum's own Statistical Data project, done on the real Palmer Penguins data.
Soil nutrients, pH and rainfall in; the crop that suits the field out — a model that asks the five most similar fields what grew well there.
A screening model for diabetes at a health camp, and why catching every case matters more than a high accuracy when a miss can hurt someone.
A regression model that predicts a home's summer electricity from what is in it — and shows exactly what each hour of AC costs.
Clustering with no labels at all: k-means finds the careful, typical and heavy water users in a housing society by itself.
Spotting the students who may need extra help early in the year — a rule the computer learns for itself, set against the attendance rule everyone assumes, and the ethics of letting a model label children.
Today's pollution and tomorrow's wind forecast in; a one-day warning out — the perceptron from Unit 2, finding its own weights.
Computer Vision
The data is images — one folder of pictures per class. Curriculum sample: Early detection of coral bleaching.
The curriculum's own Computer Vision project: reading reef photos pixel by pixel to tell healthy coral from bleached.
Counting the brown and yellow pixels on a leaf to catch crop disease early — and saying how much of the leaf is affected.
Reading an image and identifying its shape — the practical-file program, grown into a project that tells mandatory, cautionary and informatory road signs apart.
An edge-detecting kernel slides over a photo of a wall, and the cracks light up while the harmless stains do not — the convolution operator from Unit 5, doing a real job.
Natural Language Processing
The data is text — messages, reviews, questions. Curriculum sample: Sentiment analysis.
The curriculum's own NLP project: sentiment analysis of feedback on city buses — and why "no complaints" fools a list of words but not a model that learns.
A Bag of Words compares each message with the average scam and the average genuine message — and learns why a real bank alert is the hardest thing to tell apart from a fake one.
Text classification with four classes: a model reads a citizen's complaint — in English or in Hindi typed in English letters — and picks the department that should fix it.
TFIDF matches a farmer's question — in English or Hindi typed in English letters — to the right answer, and knows when to say "I'm not sure": a smart bot, built from scratch.