You've already used generative AI today. The reel captions in your feed, the "smart reply" in Gmail, the code your senior pasted from ChatGPT, the customer-care chat that answered before any human woke up — all of it is the same technology. And right now, companies are hiring people who understand how it actually works — not at PhD level, at engineer level. That's exactly the level this free course teaches, the way a friend would — with diagrams, real examples, and the actual questions companies ask at the end of every lesson.
What is Generative AI?
Generative AI is software that creates new content — text, code, images, audio — instead of just labelling existing content. Old-school AI answered questions like "is this email spam?" (yes/no). Generative AI answers "write a polite reply to this email" — and produces something that never existed before. The engines behind today's wave are LLMs (Large Language Models) — ChatGPT, Claude, Gemini — and this course is mostly about them: how they work, how to control them, and how to build real systems around them.
The one mental model that unlocks everything
Here it is, the sentence this whole field stands on: an LLM is a machine that predicts the next word (token), over and over, ridiculously well. That's it. It read a huge chunk of the internet during training, learned the patterns of language and knowledge, and now, given any text, it continues it one token at a time. Every essay, every code snippet, every chatbot answer is that one trick in a loop. Keep this sentence in your pocket — in this course we'll use it to explain hallucinations, prompting, token pricing, RAG, agents… everything.
Your first GenAI code — right now
Talking to an LLM from Python is genuinely this small. You send messages, you get a message back:
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gpt-4o-mini",
messages=[
{"role": "user",
"content": "Explain UPI to my grandmother in 2 lines."},
],
)
print(response.choices[0].message.content)Result
UPI is like sending money with a text message — you choose the person, type the amount, enter your PIN, and it reaches them instantly, any time of day.
No model training, no GPUs, no PhD — a POST request and some JSON. The engineering (and the interviews) are about everything around this call: choosing what goes in, controlling what comes out, attaching your own data, and keeping it safe. That's the course.
Why learn GenAI (the honest version)
Three reasons, no hype. One: it's the fastest-growing skill requirement in software hiring — TCS, Infosys, Accenture and Cognizant are staffing GenAI projects for global clients, and product companies expect every SDE to be able to build with LLMs. Two: the interview bar is low but real — most candidates can say "ChatGPT" but can't explain a token, RAG, or temperature. Two weeks of this course puts you ahead of that crowd. Three: it compounds — GenAI questions are appearing inside regular placement rounds too, alongside SQL and DSA, from TCS NQT to product-company system design.
How this course works
Every lesson follows the same rhythm: a real-life situation first, then the intuition, then the technical name, then working code/examples with outputs, and diagrams wherever a picture beats a paragraph. At the end of each lesson comes the Practice Zone — click-to-answer MCQs and hands-on tasks, with the companies where each pattern has been asked shown right below the question. Attempt first, reveal second. No lesson assumes any machine learning background — if you can read basic Python, you can read this course.
Course roadmap — 12 lessons
| # | Lesson | What you'll be able to answer |
|---|---|---|
| 1 | What is Generative AI? | Generative vs discriminative — the interview opener |
| 2 | How LLMs Work | Tokens, embeddings, next-token prediction |
| 3 | Transformers | Attention, encoder vs decoder — without the math |
| 4 | The LLM Landscape | GPT vs Claude vs Gemini vs Llama; open vs closed |
| 5 | LLM Settings | Temperature, top-p, tokens, context window, cost |
| 6 | Prompting Basics | Zero-shot, few-shot, chain-of-thought, system prompts |
| 7 | Hallucinations | Why models lie confidently, and how to reduce it |
| 8 | LLM APIs & Tool Calling | Chat APIs, JSON output, function calling — with code |
| 9 | RAG | Chunks, embeddings, vector DBs — the full pipeline |
| 10 | Fine-tuning vs RAG | The most-asked GenAI interview question, settled |
| 11 | AI Agents | ReAct loop, tools, memory, multi-agent systems |
| 12 | Guardrails & Safety | Prompt injection, bias, evaluation — production reality |
Company-wise GenAI PYQs
After the lessons, there's a full section of company-wise GenAI interview questions — the patterns reported from real selection rounds at TCS, Infosys, Accenture, Amazon, Microsoft and more — each one in attempt-then-reveal format, linked back to the lesson that teaches it. Treat them as mock tests once you finish the course.
How to study this course
Go in order — each lesson quietly uses the previous ones (you can't love RAG until tokens and embeddings feel obvious). Read one lesson per sitting, and always attempt the Practice Zone before revealing any answer — recalling is what makes it stick, reading is not. When a company question defeats you, follow its lesson link back, re-read that section, and return. Twelve honest sittings and you'll walk into a GenAI round knowing more than most people interviewing you.
FAQ
Do I need machine learning or heavy math background?
No. This course deliberately stays at placement/engineer level — intuition and diagrams instead of derivations. You should be comfortable reading basic Python; that's the only prerequisite.
Is this enough for GenAI interview rounds at service companies?
For the GenAI portion of freshers and early-career rounds at companies like TCS, Infosys, Accenture, Cognizant and Wipro — yes, this covers the questions that actually get asked, and each lesson's Practice Zone plus the company PYQ pages are built from those patterns. Product-company SDE rounds add system-design depth, which lessons 9–12 introduce.
Which topics matter most if I'm short on time?
Lessons 2 (how LLMs work), 9 (RAG) and 10 (fine-tuning vs RAG) — those three answer the majority of reported interview questions. Then 6 (prompting) and 11 (agents). But the short lessons are short precisely because we trimmed them, so the full course is 12 comfortable sittings, not 40.
Will this course teach me to train my own model?
No — and that's deliberate. Training models from scratch is a research-lab activity; companies hire engineers to build with existing models: prompting, APIs, RAG, agents, guardrails. That's what interviews test and what this course teaches. (Fine-tuning gets an honest, intuition-level treatment in lesson 10.)
Ready? Start here — Lesson 1: What Is Generative AI? →