Round 1 – Take-Home Assignment

The first round involved a take-home assignment that had to be completed within a few hours. The goal was to evaluate how quickly a candidate could understand and implement a new framework.

Assignment Problem Statement

Build an AI system that:

➜ Accepts a blood test report in PDF format

➜ Extracts and understands the medical data from the report

➜ Identifies possible health issues

➜ Generates suggestions for those issues

➜ Fetches suggestions from online blog articles and provides source links

The assignment tested practical application of document parsing, prompt engineering, and integrating external information sources. Even though I had not used CrewAI before, I was able to complete the assignment within the given time, which was positively noted.

Round 2 – DSA and Technical Screening

This round focused on basic programming and logical reasoning.

DSA Questions

➜ Print all prime numbers between 0 and 100

➜ Check whether two strings are anagrams of each other

These questions tested fundamentals. The anagram problem was solved using dictionaries, although it could also be solved using sorting.

General Technology Questions

➜ What is Docker?

➜ Why do we use Selenium?

➜ Have you heard about Redis?

The interviewer was checking awareness of common tools used in modern AI and backend workflows.

Round 3 – LLM and Generative AI Concepts

This round focused on core GenAI fundamentals.

GenAI Questions

➜ What is an RAG model? Explain the complete process.

➜ What are embeddings?

➜ How does chunking work in LLM-based systems?

The discussion revolved around understanding retrieval pipelines, vector representations, semantic search, and how LLMs interact with external knowledge bases.

Round 4 – Managerial Round

The managerial round focused on project experience and ownership.

Discussion Areas

➜ Detailed discussion on the take-home assignment

➜ How I approached the problem under time constraints

➜ Questions about my previous internship

➜ What level of prompts I had written in earlier projects

➜ Types of AI and GenAI projects I had worked on

The interviewer was particularly interested in adaptability and learning speed rather than prior exposure to specific frameworks.

Round 5 – Speed-Based Live Coding Test

This was a scenario-based live coding round with a strict time limit of 30 minutes.

Coding Task

➜ Given a complex JSON file, extract a specific portion of the JSON based on a pattern

➜ Pass the extracted data to an AI model

➜ Generate a summary of that data using the model

I was explicitly allowed to use external resources such as a browser or ChatGPT during this round. The task tested speed, problem-solving, and practical integration skills rather than memorization.

Final Outcome

I was selected and received the internship offer for the Generative AI Engineer role.

Key Takeaways from the Interview

Basic DSA questions are usually sufficient, especially LeetCode Easy-level problems. Strong understanding of LLM fundamentals, RAG pipelines, embeddings, chunking, and vector databases is essential. Interviewers value problem-solving approach, adaptability, and the ability to learn new tools quickly. Speed and persistence matter, especially in time-bound coding rounds.

Preparation Tips

Focus on mastering basic DSA problems. Build hands-on projects involving LLMs and document processing. Understand RAG pipelines end-to-end. Practice explaining your thought process clearly. Stay calm during live coding rounds and keep working until time runs out.

Final Thoughts

This interview experience reinforced that Generative AI interviews are increasingly practical and application-driven. You do not need to know every framework in advance, but you must demonstrate strong fundamentals, learning ability, and execution speed. Even when stuck, continuing to reason and iterate can make the difference.