HR rejection hurts more than breakups. 🥲

XYZ Campus placement Interview Experience | AI/ML Role | Rejected After HR

Recently attended XYZ's campus hiring process for the AI/ML role. Unfortunately, got rejected after reaching the final HR round. Sharing my experience for future candidates.

Recruitment Statistics

Round	Candidates
Online Assessment (OA)	250+
Technical Round 1	35
Technical Round 2	18
HR Round	7
Final Selected	6

Round 1 - Online Assessment (OA)

Duration: 90 minutes

Questions:

1. K Closest Points to a Point. (Spent Most of my time)
2. Battle Royale - Simple array and sorting problem.
3. Easy array-based question.

Solved all 3 questions within 50 minutes.

Round 2 - Technical Interview 1 (60+ Minutes)

Started with self-introduction, followed by an in-depth discussion of my AI projects.

First project:

• Which Text-to-Speech model did you use? How does it work?
• Explain Transformer architecture.
• Encoder vs Decoder architecture.
• How does the Attention mechanism work?
• Detailed questions about project implementation.

Second project (Published Research Paper):

• Explain the complete architecture and pipeline.
• How does the pipeline work?
• Can the safety mechanisms used in your research be applied to other AI systems?
• How generalizable is your approach?

Programming question:

Interviewer: Rate yourself in Python.

Me: 7.5-8/10 in Python, but 9.5-10/10 in Java since I primarily use Java for DSA.

Interviewer: "I only need Python."

He also asked about Speech-to-Text models, which I couldn't answer properly.

Answered most questions except around 3. The interview lasted over an hour, and I was shortlisted for the next round.

Round 3 - Technical Interview 2 (60+ Minutes)

Started with an introduction. The interviewer asked me to choose any AI project and draw its complete architecture on the whiteboard.

Spent considerable time explaining the architecture and pipeline.

He then asked multiple scenario-based questions. I answered most, but one scenario exposed a flaw in my architecture, and I had to adopt an alternative approach he had suggested.

Conceptual Questions:

• How does Cosine Similarity work?
• How is TF-IDF better than normal keyword-based search?
• Explain the working of semantic search and its relevance in RAG.

RAG-Based Questions:

Q1. User query: "Where does Amant live?"

• Chunk 1: "Where does Amant live? What is Amant's age? What is his current salary?"
• Chunk 2: "Bangalore, 25, 100000"

How would your RAG system retrieve and generate the correct answer?

Q2. User query: "What is XYZ CRM?"

• Chunk 1: "XYZ customer relationship management means..."
• Chunk 2: "XYZ was founded in 2011."

How would your RAG system handle this query?

Q3. RAG Debugging Scenario:

Your entire RAG pipeline is working correctly, retrieval and generation are functioning, but the system still fails to give the expected answer.

How would you identify and fix the issue?

This round also lasted over an hour.

Round 4 - HR Interview

After a long waiting period, I was finally called for HR.

Questions asked:

• Have you done any internships? Explain your work.
• Which project are you most proud of, excluding academic projects?
• Where are you from?
• What do your parents do?

I explained my internship and then she asked several follow-up questions and seemed interested. She even asked me to build something similar for her, and I mentioned that I already had a related project in my GitHub repository.

Discussed my personal project, answered a few follow-ups, and the interview ended.

Few Questions from tech round

Whats one thing you would like to automate in your life if you can? Whats the one tool that you love to use other than chatgpt and gemini? How does Cosine Similarity work? How does text to speech works? Where is QKV used in encoders and decoders model? What is BERT trained on?

Final Result - Rejected

After clearing both technical rounds and attending HR, I was quite confident.

They had printed 7 offer letters, but during the final announcement, only 6 names were called.

Unfortunately, mine wasn't among them.

Key Takeaways

• Know your projects deeply, especially architecture, implementation, and limitations.
• Prepare Transformers, Attention, Encoder-Decoder, TTS, and STT.
• Understand Cosine Similarity, TF-IDF, semantic search, and RAG.
• Practice scenario-based AI system design and RAG debugging.
• Be strong in Python for AI/ML roles.
• Reaching HR doesn't guarantee selection.