Gen AI and Agentic AI Interview Preparation Guide

 

Gen AI and Agentic AI Interview Preparation Guide — Universal

Ritesh Sinha

3 min read

·

Just now

Press enter or click to view image in full size

I am sharing topics, which are a must read before any Gen AI and Agentic AI. So lets start.

Gen AI

1. Deep Learning [ RNN, LSTM, Neural Network]
2. Transformer Architecture [Self Attention(Q,K,V concept)]
3. How does Temperature work at the core of transformer architecture?
4. RAG [ Its not only you learn high level process like loading, parsing, chunking, embedding, retrieval, prompting and generation]. You need to deep dive in each and every component. Lets see below in sub pointers.

[Loading] Adding Authentication, Guardrails, Queue management, load balancer, async processing of documents/records from the source.

[Parsing] Which parser you are going to use and how are you going to parse different element of the document.

[Chunking] How do you decide chunking strategy? Which chunk technique you use and do you use same chunking technique for every element like tables, image, scanned pages, graphs, flowchart, text? Also prepare on choosing chunk size and overlap size.

[Embedding] How do we create index, which type of metadata we shall consider, how to implement RBAC, how many index we need to create based on use cases, which embedding model we need to select and what embedding dimension is ideal.

[Retrieval] How do you implement filter and RBAC in the retrieval? Do you know how HSNW (Vector Search works)? What is role of Recall and Precision in retrieval? When do we apply reranking, when we apply hybrid search? How to evaluate retrieved context? what is RRF? What is BM25? What is TF-IDF?

[Prompting] How do you define a good optimized prompt? Types of prompting, how do you ground your prompt?

[Generation] LLM, Parameter, structured output etc.

Agentic AI

1. Agent — Types of agents, Reflection, Critic, planner, executor, ReAct, Supervisor, Tool Calling, Chain of Thought etc
2. Orchestrator : Langgraph [ StateGraph, Memory Checkpointer, FAN Out, Fan In, parallel processing, Reducers, interrupt, Resume, graph.invoke(), Seesion, thread, Memory Management.
3. Gemini ADK, Open AI SDK.

Evaluation

1. How to evaluate Gen Ai/Agentic AI system pre production(offline), After deployment (Online)
2. Evaluation of RAG, Deterministic LLM response.
3. To evaluate Agents Lifecycle (Planning, trajectory, tool calling, RAG quality, iteration, fallback hits, guardrail hits etc)
4. Structured output (

Observability

1. Application telemetry
2. Latency (Agents, Retriever, Tool communication, LLM response etc)
3. Spans of each stage
4. Input tokens, output tokens, Cost, etc.

Scenario Based

How to process millions of documents in scaled way?

How to ingest data with high scale?

How to implement RBAC based ingestion?

How to process PDF with different elements in it?

Create a chatbot for different types of doc stored in Vector DB, where millions of user does the query on it.

Python coding

1. Reverse a string
2. Fibonacci
3. Sorting of list without sort function
4. Frequency calculation in dictionary
5. Palindrome numbers
6. Finding of longest substring in a string
7. Find the elements from the list whose sum/difference is equals to target

Deployment

1. Docker/Kubernetes based deployment in any cloud.
2. Service account, API Gateway, DB management, Memory management, CI/CD.
3. Deployment through Dev, QA, Production

Project Explanation, Architecture of Project

Genai

Agentic Ai

Deployment

Gcp

Interview Questions

Written by Ritesh Sinha

31 followers

·27 following

👋 Hi, I’m Ritesh Sinha Generative AI Engineer | AI Content Creator | Helping people decode the future of artificial intelligence.