Question 1

What is Generative AI?

Accepted Answer

Generative AI is a specialized branch of Artificial Intelligence focused on creating completely new, original content—such as text, high-resolution images, synthesized audio, videos, or computer code—by learning and mimicking patterns from massive datasets. Unlike traditional AI systems that are designed to analyze existing inputs, classify data, or predict probabilities, Generative AI models are active creators. Popular foundation frameworks and applications like OpenAI's ChatGPT, Google's Gemini, Anthropic's Claude, and Midjourney leverage deep neural network structures to generate human-like contextual outputs.

Question 2

How is Generative AI different from Traditional AI?

Accepted Answer

Traditional AI focuses on analyzing, classifying, or predicting outcomes based on existing data patterns (e.g. fraud detection or spam filters). Generative AI focuses on synthesizing entirely new data samples that resemble the training data (e.g. writing articles, coding, or creating artwork). The core difference lies in their operational objectives. Traditional AI models are often 'discriminative'—they try to find decision boundaries to separate classes (e.g., Spam vs. Not Spam). Generative AI models are 'generative'—they attempt to model the actual probability distribution of the data classes to draw new instances directly from it.

Question 3

What are some real-world applications of Generative AI?

Accepted Answer

Generative AI is widely applied in corporate and creative domains, including automated customer support, semantic content generation, software code generation, medical drug discovery, high-fidelity synthetic data generation, and multimodal business assistants. Enterprise systems leverage Generative AI to boost developer velocity, write legal document drafts, and summarize multi-page financial audits. In technical QA automation, it is used to write comprehensive Cypress or Playwright test suites from standard visual mockups or functional descriptions.

Question 4

What is a Large Language Model (LLM)?

Accepted Answer

A Large Language Model (LLM) is a highly parameterized deep learning model trained on vast collections of text data utilizing the Transformer architecture to comprehend, process, and generate human language in a highly contextual manner. LLMs use statistical probability to predict the most logical next token in a sequence. By training on petabytes of text, they capture lexical patterns, grammar rules, historical facts, and reasoning shortcuts. Popular models include OpenAI's GPT, Anthropic's Claude, Google's Gemini, and open-weights models like Meta's Llama and Mistral.

Question 5

What is the difference between AI, Machine Learning, Deep Learning, and Generative AI?

Accepted Answer

AI is the broad umbrella of enabling machines to mimic intelligent human behavior. Machine Learning is a subset where models learn patterns from data. Deep Learning is a subset of ML that uses multi-layered neural networks. Generative AI is a specialized subset of Deep Learning focused on creating new content. The relationship is hierarchical, with each level representing a narrower, more specialized field of research. You can visualize the containment structure as: 
**Artificial Intelligence** ⊃ **Machine Learning** ⊃ **Deep Learning** ⊃ **Generative AI**.

Question 6

What is training data in Generative AI?

Accepted Answer

Training data is the massive, curated corpus of information (such as text, images, or code) used to teach an AI model. The model analyzes this data during the training phase to adjust its internal parameters (weights) and learn underlying statistical patterns. The capability and safety of a Generative AI model depend directly on the diversity, clean formatting, and scale of its training data. For example, language models are trained on massive datasets like Common Crawl (web pages), books, scientific papers, and open-source code repositories, while image generators are trained on paired image-text datasets.

Question 7

What is a prompt in Generative AI?

Accepted Answer

A prompt is the natural language input, instruction, question, or context provided by a user to guide a Generative AI model's response. A prompt acts as the steering wheel for an LLM or image generator. By providing clear constraints, role personas, and output requirements, the prompt guides the model's statistical generator towards producing the desired response structure.

Question 8

What is Prompt Engineering?

Accepted Answer

Prompt Engineering is the practice of designing, optimizing, and structuring prompts to guide AI models into producing highly accurate, consistent, and correctly formatted outputs. Prompt Engineering has evolved from simple text inputs into a structured methodology. It involves techniques like defining clear personas (System prompts), Zero-shot/Few-shot prompting, Chain-of-Thought (forcing step-by-step reasoning), and structured markdown layouts to prevent models from generating incorrect formatting or losing track of context.

Question 9

What are tokens in Generative AI?

Accepted Answer

Tokens are the basic building blocks of text processed by an LLM. Rather than reading full words or individual characters, models split text into small sub-word sequences called tokens, mapping them to numerical IDs before processing. Tokenization is the first step in the LLM pipeline. For example, common words might map to a single token (e.g., 'cat'), while complex or rare words are split into multiple tokens (e.g., 'auto-waiting' -> 'auto', '-', 'waiting'). As a rule of thumb, 100 English words represent approximately 130 to 140 tokens. Models have strict token limits governing the maximum size of combined prompts and responses.

Question 10

What is a Transformer model?

Accepted Answer

A Transformer is a deep learning neural network architecture introduced by Google in 2017 ('Attention Is All You Need') that relies on the **Self-Attention mechanism** to process sequence data, capturing long-range contextual relationships in parallel. Older architectures like Recurrent Neural Networks (RNNs) processed text sequentially (word-by-word), which made training slow and struggled to retain context over long distances. The Transformer processes all tokens in parallel. It uses Self-Attention to calculate mathematical weights representing the relationship between every single word and all other words in the sentence, regardless of their distance.

Question 11

Why are Transformers important in Generative AI?

Accepted Answer

Transformers revolutionized Generative AI by supporting parallel processing, which enables training on massive datasets on GPUs, and by using attention mechanisms that allow the model to retain context across long sequences of text. Because Transformers process text sequences in parallel rather than sequentially, training speed scales effectively with modern GPU clusters. Furthermore, by calculating attention weights across the entire context window, they avoid the 'forgetting' issues that plagued older RNNs, allowing modern models to analyze entire books or large codebases in a single prompt.

Question 12

What is Natural Language Processing (NLP)?

Accepted Answer

Natural Language Processing (NLP) is a broad field of Artificial Intelligence focused on enabling computers to understand, interpret, analyze, and generate human language in a structured and meaningful way. NLP combines computational linguistics, machine learning, and deep learning algorithms. It covers a wide range of standard language processing tasks, including Part-of-Speech tagging, Named Entity Recognition (NER), sentiment analysis, text classification, and machine translation.

Question 13

What is the difference between Generative AI and NLP?

Accepted Answer

NLP is the broader field of AI focused on understanding, processing, and analyzing human language. Generative AI is a specialized subset of AI focused on creating new content, which uses advanced NLP models for language tasks. NLP includes many non-creative tasks (such as sentiment analysis, entity extraction, and grammar checking) that do not generate new text. Generative AI uses advanced deep learning architectures (like Transformers) to active synthesize entirely new outputs. For example, using AI to determine if a review is positive or negative is an NLP task, while using AI to write a response to that review is a Generative AI task.

Question 14

What is fine-tuning in Generative AI?

Accepted Answer

Fine-tuning is the process of taking a pre-trained, general-purpose base model (trained on massive internet data) and training it further on a smaller, domain-specific dataset to adapt it for a particular task or industry. Base models have broad knowledge but lack specialized, proprietary information. Fine-tuning adjusts the model's internal weights using supervised learning on high-quality domain data (e.g. medical records, corporate API documentation). This helps the model adopt specific tones, follow complex formatting rules, and answer industry-specific questions with high accuracy.

Question 15

What is Retrieval-Augmented Generation (RAG)?

Accepted Answer

Retrieval-Augmented Generation (RAG) is an architectural pattern where an LLM retrieves relevant information from an external data source (like a document database) before generating its final response, ensuring accurate and up-to-date answers. LLMs are limited to the information present in their training data, which can quickly become outdated. RAG solves this by looking up relevant passages in a vector database using semantic search, inserting those passages into the prompt as context, and asking the LLM to write a response based solely on that context. This reduces hallucinations and enables real-time data access without expensive fine-tuning.

Generative AI Interview Prep Portal

What is Generative AI?

How is Generative AI different from Traditional AI?

What are some real-world applications of Generative AI?

What is a Large Language Model (LLM)?

What is the difference between AI, Machine Learning, Deep Learning, and Generative AI?

What is training data in Generative AI?

What is a prompt in Generative AI?

What is Prompt Engineering?

What are tokens in Generative AI?

What is a Transformer model?

Why are Transformers important in Generative AI?

What is Natural Language Processing (NLP)?

What is the difference between Generative AI and NLP?

What is fine-tuning in Generative AI?

What is Retrieval-Augmented Generation (RAG)?

What are embeddings in AI?

What is a Vector Database?

What are hallucinations in Generative AI?

How can hallucinations be reduced?

What is the context window in an LLM?

What is temperature in Generative AI?

What is zero-shot learning?

What is few-shot prompting?

What are AI Agents?

What are some primary challenges in Generative AI?

What is Responsible AI?

What skills are important for a Generative AI Engineer?

What is the future of Generative AI?

What is a Small Language Model (SLM)?

What are LLM Parameters and why do they matter?

How do attention mechanisms work in transformers?

What are Generative Adversarial Networks (GANs)?

What are the core components of an AI agent?

How does memory work in Agentic AI systems?

How do you implement a basic subword tokenization algorithm in Python?

Write a function to calculate the cosine similarity between two embedding vectors.

How do you implement a dynamic few-shot prompt template function in Python?

Implement a simple AI agent class with basic short-term memory functionality in Python.

How do you write a mock text generation loop demonstrating transformer autoregressive generation?

Implement a mathematical formula in Python to estimate the parameter count of a standard Transformer layer.

Generative AI
Interview Prep Portal