Ravindra BagaleCourses & study guides

Generative AI Revision Cards (Ravi Sir's Cards): Tokens, Prompts, RAG

By Last updated:

Ravi Sir's Cards for Generative AI: 3 quick revision cards and 28 comparison cards, plus 50 interview cards. Each card explains one idea in easy English with a real example. Save them on your phone or download the PNG. The QR code on each card opens its full lesson.

Revision card: Tokens and next-token prediction. Example sentence “Pune is famous for vada pav.” split into 9 tokens; the AI guesses the next token; 1 token is about three quarters of an English word.
Card 1 of 3: Tokens and next-token prediction · Download PNG · Open card page

Lesson: Tokens and Context Window Explained

Revision card: Parts of a good prompt: role, task, context and format. Before: “Write an email about leave.” After: an HR manager prompt with 18 paid leaves and a 120-word, 3-bullet format.
Card 2 of 3: Parts of a good prompt · Download PNG · Open card page

Lesson: Prompt Engineering for Beginners: Role, Task, Context and Format

Revision card: RAG in 3 steps: retrieve, augment, generate. A company-policy chatbot finds “Sick leave: 12 days per year” and answers “You get 12 sick days a year.”
Card 3 of 3: RAG in 3 steps · Download PNG · Open card page

Lesson: What Is RAG? Build a Chatbot from Your Own PDF

Comparison cards

28 comparison cards: each one puts two or more Gen AI ideas side by side, explains every word in easy English and gives a real example. The QR code on each card opens its full lesson.

Comparison card: AI vs Machine Learning vs Deep Learning vs Generative AI. AI (Artificial Intelligence): any computer system that does work which needs human-like thinking. Machine Learning (ML): a part of AI that learns patterns from examples, not from written rules. Deep Learning: a part of ML that uses neural networks (many layers of simple maths units). Generative AI: a part of Deep Learning that creates new text, images or sound.
Comparison card 1 of 28: AI vs Machine Learning vs Deep Learning vs Generative AI · Download PNG · Open card page

Lesson: Generative AI and Prompt Engineering: Course Outline

Comparison card: RAG vs Fine-tuning. RAG (Retrieval-Augmented Generation): before answering, the system finds the right parts of your documents and adds them to the prompt. Fine-tuning: you train an existing model more on your own examples, so the model itself changes.
Comparison card 2 of 28: RAG vs Fine-tuning · Download PNG · Open card page

Lesson: What Is RAG? Build a Chatbot from Your Own PDF (Step by Step)

Comparison card: Zero-shot vs Few-shot prompting. Zero-shot: you give only the instruction, with no examples. Few-shot: you give the instruction plus 2 to 5 solved examples (input → answer), so the AI copies the pattern.
Comparison card 3 of 28: Zero-shot vs Few-shot prompting · Download PNG · Open card page

Lesson: Few-Shot Prompting: Teach AI with Examples (Copy-Paste Prompts)

Comparison card: Normal software vs Generative AI. Normal software (calculator, Excel formula): follows exact rules that a programmer wrote. Generative AI (ChatGPT, Gemini): learned patterns from huge text and creates a new answer each time.
Comparison card 4 of 28: Normal software vs Generative AI · Download PNG · Open card page

Lesson: What Is Generative AI and How Does ChatGPT Work? (Simple Explanation with Examples)

Comparison card: Pre-training vs Instruction tuning. Pre-training: the first, biggest training. The model reads a huge amount of text and learns to guess the next word. Instruction tuning: later training on examples of questions with good answers, often with human feedback (people rate answers).
Comparison card 5 of 28: Pre-training vs Instruction tuning · Download PNG · Open card page

Lesson: What Is an LLM? Large Language Models Explained Simply

Comparison card: Training vs Inference. Training: building the model. It learns from huge data, and its settings (called weights) change. Inference: using the finished model. You send a prompt and it makes an answer. The weights do not change.
Comparison card 6 of 28: Training vs Inference · Download PNG · Open card page

Lesson: Generative AI and Prompt Engineering: Course Outline

Comparison card: Small, fast model vs Big, careful model. Small, fast model: answers in a second or two and costs less. Good for easy jobs. Big, careful model (often a “reasoning” or “thinking” model): takes more time to think step by step before it answers.
Comparison card 7 of 28: Small, fast model vs Big, careful model · Download PNG · Open card page

Lesson: What Is an LLM? Large Language Models Explained Simply

Comparison card: The model vs the app. Model: the AI “brain” that turns your prompt into an answer, e.g. GPT, Gemini, Claude models. App: the website or phone app you use to talk to a model, e.g. ChatGPT, Gemini app, Claude app.
Comparison card 8 of 28: The model vs the app · Download PNG · Open card page

Lesson: What Is an LLM? Large Language Models Explained Simply

Comparison card: Low temperature vs High temperature. Temperature: a setting that controls how much randomness the model uses when it picks the next word. Low (near 0): picks the most likely words. High (around 1 or more): takes more chances.
Comparison card 9 of 28: Low temperature vs High temperature · Download PNG · Open card page

Lesson: Generative AI and Prompt Engineering: Course Outline

Comparison card: Open models vs Closed models. Open model (open-weight): the company shares the model files (weights), so anyone can download and run it, e.g. Llama, Mistral, Gemma. Closed model: you can only use it through the company's app or API, e.g. GPT, Gemini, Claude.
Comparison card 10 of 28: Open models vs Closed models · Download PNG · Open card page

Lesson: Generative AI and Prompt Engineering: Course Outline

Comparison card: Weak prompt vs 4-part prompt. Weak prompt: a short request with no details. The AI must guess. 4-part prompt: gives Role (who the AI acts as), Task (the job), Context (facts it needs) and Format (shape of the answer).
Comparison card 11 of 28: Weak prompt vs 4-part prompt · Download PNG · Open card page

Lesson: Prompt Engineering for Beginners: Role, Task, Context and Format (with 5 Examples)

Comparison card: Direct answer vs Step-by-step prompt. Direct prompt: asks only for the final answer. Step-by-step prompt (chain-of-thought): asks the AI to show each step of working before the final answer.
Comparison card 12 of 28: Direct answer vs Step-by-step prompt · Download PNG · Open card page

Lesson: Step-by-Step Prompts: How to Make AI Think Before It Answers

Comparison card: Chat vs Project vs Memory in ChatGPT. Chat: one conversation. Project: a folder that keeps related chats, files and instructions together. Memory: small facts ChatGPT remembers about you across all chats.
Comparison card 13 of 28: Chat vs Project vs Memory in ChatGPT · Download PNG · Open card page

Lesson: How to Use ChatGPT Step by Step: Chats, Follow-ups, Memory and Projects

Comparison card: ChatGPT vs Gemini vs Claude vs Copilot. ChatGPT: OpenAI. Gemini: Google. Claude: Anthropic. Copilot: Microsoft. All four are AI assistants: you type or talk, and they write, summarise, explain and help with code. All have free plans (Oct 2026).
Comparison card 14 of 28: ChatGPT vs Gemini vs Claude vs Copilot · Download PNG · Open card page

Lesson: ChatGPT vs Gemini vs Claude vs Copilot: Which Free AI Tool Should You Use?

Comparison card: Free vs Paid vs Company AI plans. Free plan: no cost, for learning and light use. Paid personal plan: a monthly fee for one person. Company plan (Business or Enterprise): bought by an employer for its staff.
Comparison card 15 of 28: Free vs Paid vs Company AI plans · Download PNG · Open card page

Lesson: Generative AI and Prompt Engineering: Course Outline

Comparison card: Search answer vs AI Mode vs Deep research. All 3 let an AI read the web so the answer can be fresh. Search answer: a few quick searches, a short answer with links. AI Mode (Google): splits your question into parts, searches many at once, lets you ask follow-ups. Deep research: makes a plan, reads many pages, writes a long report.
Comparison card 16 of 28: Search answer vs AI Mode vs Deep research · Download PNG · Open card page

Lesson: AI with Web Search, Deep Research and AI Mode: When the Answer Is Fresh

Comparison card: Text fits vs Text too long for the context window. Context window: the most tokens (small pieces of words) a model can look at in one go, counting your prompt, files, the chat so far and the answer. About 1 token ≈ ¾ of an English word.
Comparison card 17 of 28: Text fits vs Text too long for the context window · Download PNG · Open card page

Lesson: Tokens and Context Window Explained: Why AI Forgets Long Chats

Comparison card: ChatGPT app vs OpenAI API. ChatGPT app: for people. You type in a website or phone app and read the reply. OpenAI API (Application Programming Interface): for programs. Your code sends text and gets text back.
Comparison card 18 of 28: ChatGPT app vs OpenAI API · Download PNG · Open card page

Lesson: Generative AI and Prompt Engineering: Course Outline

Comparison card: Text vs Table vs Checklist vs JSON output. Ways to ask the AI to shape its answer: Text: normal paragraphs. Table: rows and columns. Checklist: tick-box steps. JSON: data with named fields that a program can read.
Comparison card 19 of 28: Text vs Table vs Checklist vs JSON output · Download PNG · Open card page

Lesson: Ask AI for Tables, Checklists and JSON: Structured Output Made Easy

Comparison card: Keyword search vs Meaning search. Keyword search: finds pages that contain the same words you typed. Meaning search (semantic search): turns text into lists of numbers called embeddings, so texts with similar meaning get close numbers.
Comparison card 20 of 28: Keyword search vs Meaning search · Download PNG · Open card page

Lesson: Embeddings Explained Simply: How AI Finds Similar Meaning

Comparison card: Chroma vs pgvector vs Amazon OpenSearch. All 3 store embeddings (text turned into lists of numbers) and quickly find the closest ones. A store like this is called a vector database.
Comparison card 21 of 28: Chroma vs pgvector vs Amazon OpenSearch · Download PNG · Open card page

Lesson: Vector Databases for Beginners: Chroma, pgvector and OpenSearch

Comparison card: Chatbot vs AI workflow vs AI agent. Chatbot (generative AI): answers your message with text. You do the work. AI workflow: fixed steps you design; AI does one or more steps. AI agent (agentic AI): you give a goal and tools; the AI chooses the steps itself.
Comparison card 22 of 28: Chatbot vs AI workflow vs AI agent · Download PNG · Open card page

Lesson: What Is an AI Agent? Agentic AI vs Generative AI Explained with Examples

Comparison card: What a person sees vs What an AI agent reads. A browser agent is an AI that opens web pages and does tasks for you. A person sees only the visible page. The agent reads the page code too, including hidden text.
Comparison card 23 of 28: What a person sees vs What an AI agent reads · Download PNG · Open card page

Lesson: AI Browser Agents: Let AI Do Web Tasks Safely (2026 Guide)

Comparison card: GitHub Copilot vs Cursor vs Claude Code. All 3 are AI helpers for writing code. GitHub Copilot: works inside editors such as VS Code. Cursor: a full code editor with AI built in. Claude Code: Anthropic's coding agent you run in the terminal (and in editors).
Comparison card 24 of 28: GitHub Copilot vs Cursor vs Claude Code · Download PNG · Open card page

Lesson: AI Coding Tools Compared: GitHub Copilot, Cursor and Claude Code

Comparison card: AWS AI Practitioner vs Azure AI Fundamentals vs Google Gen AI Leader. 3 beginner AI certificates from 3 cloud companies: AWS Certified AI Practitioner (exam AIF-C01). Microsoft Azure AI Fundamentals (exam AI-901). Google Cloud Generative AI Leader.
Comparison card 25 of 28: AWS AI Practitioner vs Azure AI Fundamentals vs Google Gen AI Leader · Download PNG · Open card page

Lesson: AI Certifications Compared: AWS vs Microsoft vs Google, India 2026

Comparison card: Prompt injection vs Jailbreak. Jailbreak: a user tries to talk the AI out of its safety rules. Prompt injection: harmful instructions are hidden inside data the AI reads, like a web page, email or PDF.
Comparison card 26 of 28: Prompt injection vs Jailbreak · Download PNG · Open card page

Lesson: Generative AI and Prompt Engineering: Course Outline

Comparison card: Edit an image vs Create an image. Edit: you give the AI your own photo and ask for a change. It should keep the rest the same. Create (generate): you describe a picture in words and the AI makes a new image from nothing.
Comparison card 27 of 28: Edit an image vs Create an image · Download PNG · Open card page

Lesson: How AI Photo Editing Works: Profile Photo, Festival Look and 3D Figurine in Gemini (Nano Banana)

Comparison card: Written for the eye vs Written for the ear. For the eye: text people read on a screen. Short forms and symbols are fine. For the ear: a script that an AI voice (text-to-speech) will speak aloud. Write it the way people talk.
Comparison card 28 of 28: Written for the eye vs Written for the ear · Download PNG · Open card page

Lesson: AI Voice and Music: Text to Speech, Voice-Over and Songs (ElevenLabs, Suno)

Interview cards

50 Gen AI interview question cards: each one gives a short answer in easy English, a real example, how to say it in an interview, and a common mistake or follow-up question. The QR code on each card opens its full lesson.

Interview Q&A card: What is a token, and why does it matter? A token is a small piece of text that an LLM (Large Language Model, like ChatGPT) reads and writes: a word, part of a word, or a sign like “.”. A common English word is often 1 token; about 100 tokens ≈ 75 words .
Interview card 1 of 50: What is a token, and why does it matter? · Download PNG · Open card page

Lesson: Tokens and Context Window Explained: Why AI Forgets Long Chats

Interview Q&A card: What is RAG, and when would you use it instead of fine-tuning? RAG (Retrieval-Augmented Generation): before answering, the system searches your documents, picks the matching parts and adds them to the prompt. Fine-tuning means training the model more on your examples, so the model itself changes.
Interview card 2 of 50: What is RAG, and when would you use it instead of fine-tuning? · Download PNG · Open card page

Lesson: What Is RAG? Build a Chatbot from Your Own PDF (Step by Step)

Interview Q&A card: What is an AI hallucination, and how do you reduce it? A hallucination is when AI gives an answer that sounds sure but is false or made up. It happens because an LLM predicts likely next words; it does not check facts by itself.
Interview card 3 of 50: What is an AI hallucination, and how do you reduce it? · Download PNG · Open card page

Lesson: Why AI Hallucinates and How to Fact-Check Any AI Answer

Interview Q&A card: What is Generative AI, and how is it different from traditional AI? Generative AI creates new content: text, images, audio or code. Traditional AI mostly sorts, scores or predicts from data, for example “spam or not spam”. Both learn patterns from examples.
Interview card 4 of 50: What is Generative AI, and how is it different from traditional AI? · Download PNG · Open card page

Lesson: What Is Generative AI and How Does ChatGPT Work? (Simple Explanation with Examples)

Interview Q&A card: How does ChatGPT create an answer? ChatGPT runs on an LLM (Large Language Model). It reads your message as tokens (small text pieces) and predicts the most likely next token, again and again, until the answer is complete.
Interview card 5 of 50: How does ChatGPT create an answer? · Download PNG · Open card page

Lesson: What Is Generative AI and How Does ChatGPT Work? (Simple Explanation with Examples)

Interview Q&A card: What is an LLM, and how is it trained? An LLM (Large Language Model) is an AI model trained on huge text to predict the next token (small text piece). Training has 3 stages: pre-training on lots of text, fine-tuning on example conversations, then feedback from humans to make answers helpful and safe.
Interview card 6 of 50: What is an LLM, and how is it trained? · Download PNG · Open card page

Lesson: What Is an LLM? Large Language Models Explained Simply

Interview Q&A card: What is the difference between a model and an app? A model is the trained AI “brain” that turns input into output, like GPT or Gemini. An app is the product around it: the chat screen, login, memory, file upload, web search and safety filters.
Interview card 7 of 50: What is the difference between a model and an app? · Download PNG · Open card page

Lesson: What Is an LLM? Large Language Models Explained Simply

Interview Q&A card: What is a knowledge cutoff? A knowledge cutoff is the last date of the data a model was trained on. The model does not know events after that date, unless the app adds fresh information, for example with web search.
Interview card 8 of 50: What is a knowledge cutoff? · Download PNG · Open card page

Lesson: What Is an LLM? Large Language Models Explained Simply

Interview Q&A card: What is a context window, and why do long chats “forget”? The context window is the most text, counted in tokens (small text pieces), that a model can look at in one go: your messages, its replies and any files. Text outside the window is not seen.
Interview card 9 of 50: What is a context window, and why do long chats “forget”? · Download PNG · Open card page

Lesson: Tokens and Context Window Explained: Why AI Forgets Long Chats

Interview Q&A card: What are parameters, and is a bigger model always better? Parameters are the numbers inside a model that training adjusts; they store what it learned. Models range from a few billion to hundreds of billions of parameters. More parameters usually means more skill, but also more cost and slower answers.
Interview card 10 of 50: What are parameters, and is a bigger model always better? · Download PNG · Open card page

Lesson: What Is an LLM? Large Language Models Explained Simply

Interview Q&A card: What is temperature in an LLM? Temperature is a setting that controls how random the model’s word choices are. Low (near 0) = more focused and repeatable. High (around 1 or more) = more varied and creative, but more chance of mistakes.
Interview card 11 of 50: What is temperature in an LLM? · Download PNG · Open card page

Lesson: Your First OpenAI and Gemini API Call in Python (Free Key Setup)

Interview Q&A card: What is multimodal AI? Multimodal AI can take in or produce more than one type of data, called a modality : text, images, audio, video or PDFs. You can show it a photo and ask a question in words.
Interview card 12 of 50: What is multimodal AI? · Download PNG · Open card page

Lesson: Multimodal AI: Ask Questions About Photos, PDFs and Voice Notes

Interview Q&A card: What makes a good prompt? A prompt is the instruction you give an AI. A good one has 4 parts: Role (who the AI should act as), Task (what to do), Context (facts it needs), and Format (how the answer should look).
Interview card 13 of 50: What makes a good prompt? · Download PNG · Open card page

Lesson: Prompt Engineering for Beginners: Role, Task, Context and Format (with 5 Examples)

Interview Q&A card: What is zero-shot vs few-shot prompting? Zero-shot : you give only the instruction, no examples. Few-shot : you add 2 to 5 solved examples (input → answer) so the AI copies the pattern, labels and format.
Interview card 14 of 50: What is zero-shot vs few-shot prompting? · Download PNG · Open card page

Lesson: Few-Shot Prompting: Teach AI with Examples (Copy-Paste Prompts)

Interview Q&A card: What is step-by-step (chain-of-thought) prompting? Chain-of-thought prompting asks the AI to work through the problem in steps before giving the final answer. It helps with maths, logic and multi-step tasks.
Interview card 15 of 50: What is step-by-step (chain-of-thought) prompting? · Download PNG · Open card page

Lesson: Step-by-Step Prompts: How to Make AI Think Before It Answers

Interview Q&A card: How do you get structured output (JSON) from an LLM? Structured output means the answer comes in a fixed shape, like a table or JSON (a text format of keys and values that programs read, e.g. {"name":"Asha"}). You describe the exact fields and ask for only that format.
Interview card 16 of 50: How do you get structured output (JSON) from an LLM? · Download PNG · Open card page

Lesson: Ask AI for Tables, Checklists and JSON: Structured Output Made Easy

Interview Q&A card: What is a system prompt? A system prompt is a hidden instruction that is set before the user types. It fixes the AI’s role, rules, tone and limits for the whole chat. The user’s message comes after it.
Interview card 17 of 50: What is a system prompt? · Download PNG · Open card page

Lesson: Save Your Best Prompts: Custom GPTs, Gemini Skills and Claude Projects

Interview Q&A card: What is prompt injection? Prompt injection is an attack where someone hides instructions in text the AI reads, so it ignores its real rules. It can be typed directly or hidden in a web page, email or file.
Interview card 18 of 50: What is prompt injection? · Download PNG · Open card page

Lesson: Build Your First AI Agent in Python: Tools, Memory, Guardrails

Interview Q&A card: How do you test or evaluate a prompt? Make a small test set : 10 to 50 real inputs with the answer you expect. Run the prompt on all of them, score each result with a rubric (a list of checks, like correct, complete, right format), and compare versions.
Interview card 19 of 50: How do you test or evaluate a prompt? · Download PNG · Open card page

Lesson: AI Mock Interview: Practise HR and Technical Questions with ChatGPT

Interview Q&A card: How do you fact-check an AI answer? Check the claims that matter against primary sources (the original: official site, law text, research paper). Open every link the AI gives, search key names and numbers yourself, and be extra careful with dates, prices and quotes.
Interview card 20 of 50: How do you fact-check an AI answer? · Download PNG · Open card page

Lesson: Why AI Hallucinates and How to Fact-Check Any AI Answer

Interview Q&A card: What are embeddings? An embedding is a list of numbers (a vector ) that represents the meaning of a text, image or sound. Texts with similar meaning get similar numbers, so a computer can find “close” meanings fast.
Interview card 21 of 50: What are embeddings? · Download PNG · Open card page

Lesson: Embeddings Explained Simply: How AI Finds Similar Meaning

Interview Q&A card: What is cosine similarity? Cosine similarity measures how close two embeddings (number lists that stand for meaning) point in the same direction. Near 1 = very similar meaning; near 0 = unrelated.
Interview card 22 of 50: What is cosine similarity? · Download PNG · Open card page

Lesson: Embeddings Explained Simply: How AI Finds Similar Meaning

Interview Q&A card: Fine-tuning, RAG or prompting: how do you choose? Prompting : write better instructions; cheapest, try first. RAG (Retrieval-Augmented Generation): add your own documents to each prompt; for facts that change. Fine-tuning : train the model more on examples; for a fixed style or format.
Interview card 23 of 50: Fine-tuning, RAG or prompting: how do you choose? · Download PNG · Open card page

Lesson: Fine-Tuning vs RAG vs Prompting: Which One Do You Need?

Interview Q&A card: What is fine-tuning? Fine-tuning is extra training of an existing model on your own examples (input → ideal output). The model’s internal numbers change, so it learns a style, format or narrow task.
Interview card 24 of 50: What is fine-tuning? · Download PNG · Open card page

Lesson: Fine-Tuning vs RAG vs Prompting: Which One Do You Need?

Interview Q&A card: Are AI detectors reliable? An AI detector guesses whether text was written by AI from patterns like predictable word choice. It gives a probability, not proof, and makes mistakes both ways.
Interview card 25 of 50: Are AI detectors reliable? · Download PNG · Open card page

Lesson: AI Detectors and Humanize AI Tools: What Works and What Is Honest

Interview Q&A card: What are the steps of RAG? RAG (Retrieval-Augmented Generation) has 3 steps: Retrieve the matching parts of your documents, Augment (add them to the prompt), and Generate the answer from them. Before that, documents are split into chunks and indexed once.
Interview card 26 of 50: What are the steps of RAG? · Download PNG · Open card page

Lesson: What Is RAG? Build a Chatbot from Your Own PDF (Step by Step)

Interview Q&A card: Why do we chunk documents, and what chunk size do you use? Chunking means splitting long documents into small pieces before indexing. Small pieces match questions more precisely and fit in the context window (how much text the model can read at once).
Interview card 27 of 50: Why do we chunk documents, and what chunk size do you use? · Download PNG · Open card page

Lesson: RAG on AWS: Bedrock Knowledge Bases with S3 (Step by Step)

Interview Q&A card: What is a vector database? A vector database stores embeddings (number lists that represent meaning) and quickly finds the ones closest to a question’s embedding. It powers search by meaning and the retrieval step of RAG.
Interview card 28 of 50: What is a vector database? · Download PNG · Open card page

Lesson: Vector Databases for Beginners: Chroma, pgvector and OpenSearch

Interview Q&A card: How do you know a RAG chatbot is good? Test it on a fixed set of real questions with known answers. Check 2 things separately: did it retrieve the right chunk, and is the answer correct, complete and backed by the cited source?
Interview card 29 of 50: How do you know a RAG chatbot is good? · Download PNG · Open card page

Lesson: What Is RAG? Build a Chatbot from Your Own PDF (Step by Step)

Interview Q&A card: What should a RAG bot do if the answer is not in the documents? It should clearly say it could not find the answer and point to a person or page, not guess. You set this rule in the system prompt (the fixed hidden instructions) and test it.
Interview card 30 of 50: What should a RAG bot do if the answer is not in the documents? · Download PNG · Open card page

Lesson: What Is RAG? Build a Chatbot from Your Own PDF (Step by Step)

Interview Q&A card: How do you build RAG on AWS? Put documents in Amazon S3 (file storage). Create an Amazon Bedrock Knowledge Base that chunks them, turns them into embeddings (meaning as numbers) and stores them in a vector store. Then ask questions; Bedrock retrieves and answers with citations.
Interview card 31 of 50: How do you build RAG on AWS? · Download PNG · Open card page

Lesson: RAG on AWS: Bedrock Knowledge Bases with S3 (Step by Step)

Interview Q&A card: What is an AI agent? An AI agent is an AI system that works toward a goal by itself: it plans steps, uses tools (like search, email or a calendar), checks results, and repeats until done. A chatbot only replies.
Interview card 32 of 50: What is an AI agent? · Download PNG · Open card page

Lesson: What Is an AI Agent? Agentic AI vs Generative AI Explained with Examples

Interview Q&A card: What is the difference between a chatbot and an AI agent? A chatbot answers your message and stops. An AI agent takes a goal, decides steps, uses tools (apps or APIs it can call) and acts, often across many steps, until the job is done.
Interview card 33 of 50: What is the difference between a chatbot and an AI agent? · Download PNG · Open card page

Lesson: What Is an AI Agent? Agentic AI vs Generative AI Explained with Examples

Interview Q&A card: What is tool calling (function calling)? Tool calling lets an LLM ask your program to run a function, like get_weather(city). The model only writes the request in a structured form; your code runs the tool and sends the result back for the final answer.
Interview card 34 of 50: What is tool calling (function calling)? · Download PNG · Open card page

Lesson: Build Your First AI Agent in Python: Tools, Memory, Guardrails

Interview Q&A card: What guardrails does an AI agent need? Guardrails are limits that keep an agent safe: least-privilege access (only the tools and data it needs), step and cost limits, input/output checks, logs, and human approval before risky actions like paying, deleting or sending.
Interview card 35 of 50: What guardrails does an AI agent need? · Download PNG · Open card page

Lesson: Build Your First AI Agent in Python: Tools, Memory, Guardrails

Interview Q&A card: What is MCP (Model Context Protocol)? MCP is an open standard, started by Anthropic in 2024, for connecting AI apps to tools and data. A tool is wrapped once as an MCP server ; any AI app that supports MCP (the client ) can use it. Think of it like a USB-C port for AI.
Interview card 36 of 50: What is MCP (Model Context Protocol)? · Download PNG · Open card page

Lesson: What Is MCP (Model Context Protocol)? Connect AI to Your Tools

Interview Q&A card: When would you use n8n instead of writing code? n8n is a workflow automation tool where you connect apps by dragging blocks (called nodes), with AI agent blocks built in. Use it for quick business automations; write code when you need complex logic, heavy testing or large scale.
Interview card 37 of 50: When would you use n8n instead of writing code? · Download PNG · Open card page

Lesson: Build an AI Agent in n8n Without Coding (Free, Step by Step)

Interview Q&A card: What should never go into a public AI tool? Never paste secrets (passwords, API keys, OTPs), personal data (Aadhaar, PAN, bank details, health records), or confidential company data (source code, client lists, unreleased plans) into a public AI chat unless your company has approved it.
Interview card 38 of 50: What should never go into a public AI tool? · Download PNG · Open card page

Lesson: Is ChatGPT Safe? Privacy Settings and What Never to Paste into AI

Interview Q&A card: How do you handle private data in an AI app? Collect only what you need, mask or remove personal details before sending text to a model, use business plans or APIs that don’t train on your data, control who can see logs, and follow laws like India’s DPDP Act 2023 (Digital Personal Data Protection).
Interview card 39 of 50: How do you handle private data in an AI app? · Download PNG · Open card page

Lesson: Is ChatGPT Safe? Privacy Settings and What Never to Paste into AI

Interview Q&A card: What is a deepfake, and how do you check one? A deepfake is a fake video, photo or voice made with AI to look or sound like a real person. Check by: verifying through a second channel, looking for odd lips, blinking or lighting, finding the original source, and checking content labels.
Interview card 40 of 50: What is a deepfake, and how do you check one? · Download PNG · Open card page

Lesson: Deepfakes Explained: How They Are Made and How to Check a Suspicious Video, Photo or Voice Note

Interview Q&A card: Who owns AI-generated content? It depends on the country and on how much a human created. The US Copyright Office says purely AI-made work can’t be copyrighted; human-made parts can be. Indian law has no clear rule yet. Always check the tool’s terms of use.
Interview card 41 of 50: Who owns AI-generated content? · Download PNG · Open card page

Lesson: Copyright, Bias and Fair Use of AI Images and Text

Interview Q&A card: What is bias in AI, and how do you reduce it? Bias is when an AI treats groups unfairly or repeats stereotypes, because its training data or design was unbalanced. Reduce it with diverse data, testing outputs across groups, clear rules and human review.
Interview card 42 of 50: What is bias in AI, and how do you reduce it? · Download PNG · Open card page

Lesson: Copyright, Bias and Fair Use of AI Images and Text

Interview Q&A card: What is Amazon Bedrock? Amazon Bedrock is an AWS service that lets you use many foundation models (large pre-trained AI models) from Amazon, Anthropic, Meta, Mistral and others through one API (a program-to-program interface), without managing servers.
Interview card 43 of 50: What is Amazon Bedrock? · Download PNG · Open card page

Lesson: Amazon Bedrock for Beginners: Your First Gen AI App on AWS

Interview Q&A card: How do you control the cost of a Gen AI app? Cost mostly = tokens (text pieces) in and out × price per token. Control it by picking the smallest model that passes tests, keeping prompts short, limiting output length, caching repeated answers, and setting budgets and alerts.
Interview card 44 of 50: How do you control the cost of a Gen AI app? · Download PNG · Open card page

Lesson: Amazon Bedrock for Beginners: Your First Gen AI App on AWS

Interview Q&A card: How do you keep an API key safe? An API key is a secret password that lets a program use a paid service. Keep it out of code: store it in an environment variable or a secrets manager, never commit it to GitHub, give it the least access, and rotate (replace) it if exposed.
Interview card 45 of 50: How do you keep an API key safe? · Download PNG · Open card page

Lesson: Your First OpenAI and Gemini API Call in Python (Free Key Setup)

Interview Q&A card: How do you evaluate a Gen AI feature before launch? Define what “good” means, build a test set of real cases, measure quality (correct, grounded, safe, right format), plus latency (response time) and cost per request. Then run a small pilot with real users and monitor.
Interview card 46 of 50: How do you evaluate a Gen AI feature before launch? · Download PNG · Open card page

Lesson: 50 Generative AI Interview Questions and Answers (Beginner to Practitioner)

Interview Q&A card: How does AI web search (or Deep Research) work? The AI turns your question into search queries, reads the top pages, and writes an answer with links. Deep Research does this many times in a row, over several minutes, and writes a longer cited report.
Interview card 47 of 50: How does AI web search (or Deep Research) work? · Download PNG · Open card page

Lesson: AI with Web Search, Deep Research and AI Mode: When the Answer Is Fresh

Interview Q&A card: How do AI coding assistants help developers, and what are the risks? AI coding tools give 3 kinds of help: autocomplete (suggests the next lines), chat (explains or writes code on request) and agents (make multi-file changes and run tests). Risks: wrong code, security holes and leaked secrets.
Interview card 48 of 50: How do AI coding assistants help developers, and what are the risks? · Download PNG · Open card page

Lesson: AI Coding Tools Compared: GitHub Copilot, Cursor and Claude Code

Interview Q&A card: What does a Gen AI engineer do, and how do you show you are ready? A Gen AI engineer builds apps on top of LLMs: prompts, RAG (answers from company documents), tool calling, evaluation, cost control and safety. Show readiness with small, working projects on GitHub, each with a clear README.
Interview card 49 of 50: What does a Gen AI engineer do, and how do you show you are ready? · Download PNG · Open card page

Lesson: AI Jobs in India for Freshers: Roles, Skills and 3 Portfolio Projects

Interview Q&A card: What is a foundation model? A foundation model is a large AI model pre-trained on very broad data, which can then be adapted to many tasks with prompts, RAG or fine-tuning. LLMs are one kind; image and multimodal models are others.
Interview card 50 of 50: What is a foundation model? · Download PNG · Open card page

Lesson: AWS AI Practitioner (AIF-C01): 20 Practice Questions by Domain

Back to the Generative AI course