End-of-Year Reunion — All Characters, One Celebration
It's end of year. Arjun (Visakhapatnam) built a neural net. Priya (Pune) explored her first real dataset. Kavita (Chennai) cleaned and prepared messy data. Rohan (Hyderabad) trained a Decision Tree classifier. Divya (Coimbatore) evaluated models with precision and recall. Ananya (Bengaluru) built a linear regression predictor. Aryan (Mumbai) dove into Generative AI. Sneha (Nagpur) processed language with NLP. Rahul (Jaipur) analysed data with pandas. Aisha (Lucknow) mapped her AI career path. Dev (Pune) called his first AI API.
That's 11 students across 12 lessons, in 10 Indian cities, each building something real. Now it's your turn to look back at what YOU have built, document it, and claim your certificate.
12-Lesson Recap
Everything You Covered This Year
L12
Capstone Portfolio ← You Are Here
You · Your City 🌟
Skills Checklist
Tick Everything You Can Do
Check off each skill you've genuinely practised:
0 of 12 skills checked
Portfolio Builder
Reflect on Your Best Work
Write your answers below. These form the basis of your AI portfolio. Copy them into a Google Doc and share the link on GitHub or with your teacher.
Your AI Portfolio — Class 9
1. Best Classifier Project (Lesson 04 or 05)
2. Best Data Analysis Notebook (Lesson 02, 03, or 09)
3. API Tool Project (Lesson 11)
4. Class 9 Reflection — What Changed?
GitHub tip: Create a public repository called
class-9-ai-portfolio, add all your Colab notebooks, and paste your portfolio answers in the README. This is your first real tech portfolio — it exists forever and colleges/employers can see it.
Your Certificate
Print Your Completion Certificate
First pass the Capstone Quiz below (80% or higher). The Generate button unlocks once you pass.
Mitra AI Life Education
🏆
This is to certify that
has successfully completed
Class 9 — Building with AI
12 Lessons · Neural Networks · Datasets · Classifiers · Regression · Generative AI · NLP · Data Analysis · AI Careers · API Tools
What's Next
Class 10 — Advanced AI Projects Preview
Advanced Deep Learning
CNNs, RNNs, LSTM networks, and transfer learning. Build models that go beyond what Class 9 covered.
Computer Vision Projects
Object detection, image classification, face recognition. Train a model to recognise Indian vegetables.
Advanced NLP Projects
Build a chatbot, fine-tune a language model, and work with Indian language datasets from AI4Bharat.
Full ML Pipeline
End-to-end: data collection → preprocessing → training → evaluation → versioning with DVC and MLflow.
Deploying AI Apps
Deploy a real ML model as a web app using FastAPI + Streamlit. Share a URL people can actually use.
Class 10 Capstone
Build a complete AI product that solves a real Indian problem — from idea to deployed URL.
Lesson Wisdom
One Key Insight From Each Lesson
| Lesson | Key Insight |
|---|---|
| L01 Neural Networks | A neural network learns by adjusting weights — tiny numbers that the model updates thousands of times until it gets things right. |
| L02 Datasets | Garbage in, garbage out. The quality of your data determines the ceiling of your model's performance. |
| L03 Data Cleaning | 80% of real AI work is cleaning data. The "boring" step is actually the most important one. |
| L04 Classifiers | A Decision Tree makes decisions the way humans do — splitting on one question at a time until it reaches an answer. |
| L05 Evaluation | Accuracy alone lies. A model that predicts "no disease" for everyone looks 99% accurate if only 1% of people are sick. |
| L06 Regression | Not every AI problem is a yes/no question. When you need to predict a number, regression is your tool. |
| L07 Generative AI | Generative AI doesn't "understand" — it predicts what token comes next, trained on vast human knowledge. That's both powerful and limited. |
| L08 NLP | Language is the hardest thing to teach a machine — because words change meaning based on context, culture, and tone. |
| L09 Pandas & Charts | A chart you made from real data tells a story no text paragraph can. Visualisation is how data becomes insight. |
| L10 AI Careers | The most valuable people in the AI economy combine domain expertise (medicine, law, farming) with AI tool fluency — not just one or the other. |
| L11 APIs | You don't need to train AI to use AI. An API call lets you access the world's most powerful models in 5 lines of code. |
| L12 Capstone | Building something real — even something small — teaches more than reading a hundred tutorials. Start building today. |
🏆 Final Capstone Quiz — 10 Questions Across All 12 Lessons
1. In a neural network, backpropagation is the process of:
a) Adding more layers to improve accuracy
b) Calculating how wrong the prediction was and adjusting weights backwards through the network to reduce the error
c) Feeding the training data into the input layer
d) Encoding categorical variables as numbers
2. When your model gets 98% accuracy on training data but only 61% on test data, this is called:
a) Underfitting — the model is too simple
b) A correctly trained model — test accuracy is always lower
c) Overfitting — the model memorised training data but fails to generalise
d) A data cleaning problem — dirty data caused the gap
3. In a classification problem where missing diseased patients is dangerous, which metric is most important?
a) Accuracy — highest overall correctness
b) Precision — avoid false alarms
c) Recall — minimise false negatives (missing sick patients)
d) R² — measures how well the regression line fits
4. Linear regression differs from classification because:
a) It uses more data
b) It predicts a continuous number rather than a category
c) It requires a neural network to work
d) It only works on image data
5. A GAN (Generative Adversarial Network) consists of:
a) An encoder and a decoder
b) A generator that creates fake data and a discriminator that tries to detect the fakes — they train together in competition
c) A tokeniser and a transformer
d) A classifier and a regressor
6. TF-IDF in NLP gives a word a HIGH score when it:
a) Appears frequently in all documents in the corpus
b) Is very short (one or two letters)
c) Appears often in one document but rarely across the full dataset — making it distinctive for that document
d) Is listed in the stop words list
7. In pandas, df.groupby('City')['Marks'].mean() will:
a) Sort the dataframe by city name
b) Delete all rows where City is missing
c) Calculate the average marks for students in each city — one row per city in the result
d) Filter the dataframe to show only rows where City equals Marks
8. When calling the Gemini API from Colab, a 403 error most likely means:
a) The server is temporarily too busy (rate limit)
b) Your API key is invalid, expired, or missing — you don't have permission
c) Your prompt contains more than 1000 tokens
d) The internet connection is down
9. The best way to store your API key in Google Colab is:
a) Write it directly in the code as a string variable
b) Add it in a comment so it's easy to find
c) Use Colab Secrets (🔑 icon) and access it with userdata.get() — key stays private even when you share the notebook
d) Email it to yourself for safekeeping
10. By completing Class 9 Building with AI, you now understand that AI is:
a) Magic — impossible to understand without a PhD
b) Dangerous — best avoided until fully regulated
c) A learnable set of tools and ideas — patterns from data, models you can build and evaluate, APIs you can call, and a career landscape you can navigate
d) Fully automated — computers will do all the work so humans don't need to learn it