Generative AI & Agentic AI Interview Questions and Answers
1. Generative AI Fundamentals
What is Generative AI?
Generative AI is a branch of Artificial Intelligence that creates new content such as text, images, audio, video, and code by learning patterns from training data.
How is Generative AI different from Traditional AI?
Traditional AI focuses on prediction, classification, and decision-making, while Generative AI creates new content based on learned patterns.
What is an LLM?
A Large Language Model (LLM) is a neural network trained on massive text datasets to understand and generate human-like language.
What are examples of LLMs?
- GPT-4
- GPT-4o
- Claude
- Gemini
- Llama
- Mistral
What is a Transformer Model?
Transformer is a deep learning architecture that uses self-attention mechanisms to understand relationships between words and generate context-aware outputs.
What is Self-Attention?
Self-Attention allows a model to identify important words in a sentence and understand relationships between them.
What are Tokens?
Tokens are small pieces of text processed by an LLM. They may be words, characters, or subwords.
What is Context Window?
The Context Window represents the amount of information an LLM can process in a single request.
What are Parameters in LLMs?
Parameters are learned weights inside a model that store knowledge acquired during training.
What is Pre-Training?
Pre-Training is the initial stage where a model learns language patterns from large datasets.
What is Fine-Tuning?
Fine-Tuning adapts a pre-trained model for specific business or domain requirements using specialized datasets.
What is Instruction Tuning?
Instruction Tuning trains a model to follow user instructions more effectively.
What is RLHF?
Reinforcement Learning from Human Feedback (RLHF) improves model responses based on human evaluations and rankings.
What is Hallucination?
Hallucination occurs when an AI model generates incorrect, misleading, or fabricated information while appearing confident.
How can Hallucinations be reduced?
- RAG
- Grounded Data Sources
- Validation Layers
- Human Review
- Prompt Engineering
2. Prompt Engineering
What is Prompt Engineering?
Prompt Engineering is the practice of designing prompts that improve accuracy, relevance, and consistency of AI responses.
What is Zero-Shot Prompting?
The model performs a task without examples.
What is One-Shot Prompting?
The model receives one example before solving the task.
What is Few-Shot Prompting?
The model is given several examples before performing the requested task.
What is Chain-of-Thought Prompting?
A prompting technique that encourages the model to reason step-by-step before providing an answer.
What is Role-Based Prompting?
Assigning a specific role to the model, such as Software Architect, Doctor, Teacher, or Data Scientist.
What is Prompt Injection?
A security attack where users attempt to manipulate AI instructions to bypass safety or system constraints.
How do you prevent Prompt Injection?
- Input Validation
- Content Filtering
- System Prompt Protection
- Access Control
- Human Review
3. Embeddings and Vector Databases
What are Embeddings?
Embeddings are numerical vector representations of data that capture semantic meaning.
Why are Embeddings important?
- Semantic Search
- Recommendation Systems
- Document Retrieval
- Similarity Search
- RAG Applications
What is a Vector Database?
A Vector Database stores embeddings and enables efficient similarity searches.
Name popular Vector Databases.
- Azure AI Search
- Pinecone
- Milvus
- Weaviate
- FAISS
- ChromaDB
What is Similarity Search?
Similarity Search finds the most relevant records by comparing vector embeddings.
What is Cosine Similarity?
A metric used to measure similarity between vector embeddings by calculating the angle between vectors.
What is Vector Indexing?
Vector Indexing improves search performance by organizing embeddings for faster retrieval.
4. RAG (Retrieval-Augmented Generation)
What is RAG?
RAG combines information retrieval with LLMs to generate accurate and context-aware responses.
Why is RAG important?
- Reduces Hallucinations
- Uses Real-Time Data
- Improves Accuracy
- No Model Retraining
What are the main components of RAG?
- Documents
- Chunking
- Embeddings
- Vector Database
- Retriever
- LLM
What is Chunking?
Chunking breaks large documents into small pieces before embedding.
What is Semantic Search?
Semantic Search retrieves content based on meaning rather than exact keywords.
What is Hybrid Search?
Hybrid Search combines semantic search and keyword search for better relevance.
How does a RAG pipeline work?
User Query
↓
Embedding Model
↓
Vector Search
↓
Relevant Documents
↓
LLM
↓
Response
What is Re-ranking?
Re-ranking improves retrieval accuracy by sorting retrieved documents according to relevance.
5. Agentic AI
What is Agentic AI?
Agentic AI refers to systems capable of planning, reasoning, executing tasks, using tools, and achieving goals autonomously.
What is an AI Agent?
An AI Agent is an intelligent system that perceives information, plans actions, executes tasks, and evaluates outcomes.
How is Agentic AI different from Generative AI?
Generative AI creates content, while Agentic AI makes decisions and performs actions to achieve objectives.
What are the core components of an AI Agent?
- LLM
- Memory
- Reasoning
- Planning
- Tools
- Execution Framework
What is Agent Memory?
Memory allows agents to remember interactions, user preferences, and previous tasks.
What is Long-Term Memory?
Long-Term Memory stores information that can be reused across multiple conversations or sessions.
What is Short-Term Memory?
Short-Term Memory stores the current conversation context.
What is Planning in Agentic AI?
Planning breaks complex objectives into smaller executable steps.
What is Reflection?
Reflection allows agents to analyze previous outputs and improve future decisions.
What is Tool Calling?
Tool Calling enables agents to invoke APIs, databases, search engines, and business applications.
What is Function Calling?
Function Calling enables an LLM to invoke predefined backend functions using structured parameters.
What is a Multi-Agent System?
A Multi-Agent System consists of multiple specialized agents collaborating to solve tasks.
What is Agent Orchestration?
Agent Orchestration coordinates multiple agents and workflows for complex business processes.
6. LangChain
What is LangChain?
LangChain is an open-source framework for building LLM-powered applications.
What are Chains in LangChain?
Chains combine multiple LLM operations into a workflow.
What are Agents in LangChain?
Agents dynamically select tools and actions based on user objectives.
What is LangGraph?
LangGraph is an advanced framework for building stateful multi-agent applications.
What is Memory in LangChain?
Memory enables conversational continuity across interactions.
7. Semantic Kernel
What is Semantic Kernel?
Semantic Kernel is Microsoft's framework for AI orchestration and Agentic AI development.
What are Plugins in Semantic Kernel?
Plugins expose functions that AI agents can invoke to perform actions.
What are Native Functions?
Native Functions are C# or Python methods exposed as AI capabilities.
What is a Planner in Semantic Kernel?
A Planner automatically generates execution plans to achieve user goals.
8. Azure OpenAI
What is Azure OpenAI Service?
Azure OpenAI provides enterprise access to OpenAI models with Microsoft's security and governance features.
Why choose Azure OpenAI?
- Enterprise Security
- Compliance
- Private Networking
- Scalability
- Responsible AI Controls
What is Azure AI Search?
Azure AI Search is a search platform that supports keyword, vector, and hybrid search.
What is Responsible AI?
Responsible AI ensures systems are fair, transparent, secure, and accountable.
9. Real-World Implementation Questions
How would you build an Enterprise AI Chatbot?
- Collect knowledge documents
- Perform chunking
- Create embeddings
- Store embeddings in Vector DB
- Implement retrieval layer
- Integrate LLM
- Add citations
- Deploy application
How would you reduce AI application costs?
- Caching
- Smaller Models
- Efficient Prompts
- Token Optimization
- Request Batching
How would you secure an AI application?
- Authentication
- Authorization
- Prompt Protection
- Encryption
- Logging
- Rate Limiting
How would you monitor AI systems?
- Latency Metrics
- Token Usage
- Error Rates
- Hallucination Tracking
- User Feedback
10. Advanced AI Architecture Questions
What is Model Distillation?
Model Distillation transfers knowledge from a large model into a smaller and faster model.
What is Quantization?
Quantization reduces model size by lowering numerical precision.
What is Knowledge Grounding?
Knowledge Grounding connects responses to trusted data sources.
What is MCP (Model Context Protocol)?
MCP is a standardized protocol that connects AI models with external tools, systems, and data sources.
What is an AI Copilot?
An AI Copilot is an assistant that augments human productivity by providing suggestions and automation capabilities.
What are Enterprise AI Challenges?
- Data Privacy
- Hallucinations
- Governance
- Cost Management
- Security
- Compliance
- Scalability
11. Quick-Fire Questions
- What is LLM? Large Language Model.
- What is RAG? Retrieval-Augmented Generation.
- What is an Embedding? Numerical representation of meaning.
- What is a Vector Database? Database optimized for vector storage and retrieval.
- What is Hallucination? Incorrect AI-generated information.
- What is Prompt Engineering? Designing effective prompts.
- What is Agentic AI? Autonomous AI capable of action.
- What is Function Calling? Invoking external functions from an LLM.
- What is Semantic Search? Search based on meaning.
- What is Fine-Tuning? Additional training on domain-specific data.
- What is Azure AI Search? Microsoft search and vector platform.
- What is LangChain? Framework for LLM applications.
- What is Semantic Kernel? Microsoft's AI orchestration framework.
- What is GPT? Generative Pre-trained Transformer.
- What is RLHF? Reinforcement Learning from Human Feedback.