𝗠𝘆 𝗜𝗻𝘁𝗲𝗿𝘃𝗶𝗲𝘄 𝗘𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲 𝗮𝘁 Cynlr | 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿-𝗔𝗜/𝗠𝗟

The overall process consisted of an Online Assessment followed by multiple interview rounds:

• Online Assessment
• Round 1: Group Discussion
• Round 2: Technical Round — 1
• Round 3: Domain-Specific Round — AI/ML
• Round 4: Interview with Founders

Here is my complete experience.

Online Assessment

The Online Assessment was of 3 hours duration.

The assessment consisted of different sections covering Software/Algorithmic questions, Mathematics and Physics, AI/ML, programming, and system-level concepts.

1. Software / Algorithmic Section

There was a Software/Algorithmic section containing algorithmic questions.

2. Mathematics & Physics

There were Mathematics and Physics questions in the assessment, with many of them being Physics-based.
The section consisted of both:

• MCQ questions
• Descriptive questions

3. AI/ML Questions

There were also AI/ML questions presented in the form of paragraphs, followed by questions asking us to provide our opinions.

4. Programming Questions

There were programming-related questions covering different areas.

Some questions involved calculating the memory size of a particular program.

For example, questions required calculating how many bytes a program would take based on the data types used.

There were also programming questions involving:

• Arithmetic operations
• Finding the output of a program

5. System Architecture Questions

There were also questions related to system architecture and concepts such as threads.

Overall, the Online Assessment tested my understanding of system-level thinking, basic Physics concepts, programming, Mathematics, and AI/ML, along with first-principles thinking.

Round 1: Group Discussion

The first interview round was a Group Discussion (GD).
The discussion topics included:

• Mathematics & Physics concepts
• “Jugad”

The round involved discussing the given topics with the other participants.

Round 2: Technical Round

The interviewer initially started with an introduction and then moved to my submissions from the Online Assessment.

The interviewer selected one question from my assessment and asked me to solve it and explain how I approached the problem.

Physics Problem — Equilibrium Pressure in a Water Pipeline

The question I received was a Physics-based problem related to equilibrium pressure in a water pipeline.

I explained my thought process and approach to solving the problem.

The interviewer then asked multiple follow-up questions based on the same problem.

After discussing the Physics problem, the interviewer asked me about the subjects I had learned during my B.Tech course.

I mentioned the subjects I had studied from my first semester to the final semester.

The interviewer then moved to Computer Architecture & Organisation and Compiler Construction.

Question 1: What are the types of computer architecture?

I answered with:

• Harvard architecture
• Layman architecture

Question 2: What is the difference between these architectures?

I answered that the architectures are differentiated based on how memory is organized.

Question 3: How are they actually different in terms of memory?

I explained that one architecture uses a single memory for both data and instructions, while another uses separate memories for data and instructions.

Question 4: Which architecture uses single memory and which uses separate memories?

The interviewer wanted me to be more specific about which architecture uses which memory organization.

At that point, I said that I didn’t remember exactly which memory organization belonged to which architecture, but I knew that the main difference was based on the memory organization.

The interviewer then moved to the next topic.

Question 5: What is the difference between a program and a process?

My answer was:
A program is a set of instructions, whereas a process is a program under execution.

Question 6: How does compilation work? What is the difference between a compiler and an interpreter?

The interviewer then asked about the compilation process and the difference between a compiler and an interpreter.

Question 7: What is a stack and a queue?

The interviewer asked me to explain the concepts of stack and queue.

Question 8: Implement a stack in your preferred language.

I implemented the stack using C++ and used a vector to implement it.
I implemented the basic operations including:

• Push
• Pop

After this, the interviewer asked if I had any questions.

Since Cynlr is a robotics company, I asked some company-specific questions, including:

• How do they build their robots?
• Do they build them from scratch or use pre-built components?
• What products are they building?

Towards the end, the interviewer asked me which domain I was interested in.
I said that I was interested in AI/ML.
The interviewer then told me that there would be another domain-specific round.

Round 3: Domain-Specific Round — AI/ML

The third round was the AI/ML domain-specific interview.

The interviewer started with an introduction and then asked me about the AI/ML topics I had learned and how I had learned them.

I explained that I had taken a course on ML with Python, where I learned about:

• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning algorithms
• Deep Learning
• Computer Vision concepts and algorithms

I also mentioned learning through YouTube playlists.
The interviewer then started asking AI/ML questions.

Question 1: What are the different types of supervised learning algorithms?

I answered that supervised learning includes:

• Regression
• Classification

Question 2: Why is Logistic Regression called regression even though it is a classification algorithm?

My answer was that Logistic Regression computes a linear combination of the inputs.
For the output, it uses a probability function, specifically the sigmoid function.
Then, based on a threshold, we classify the output into different classes.

So, although it is used for classification, it is called Logistic Regression because of the way the underlying model computes the output.

Question 3: What are the different activation functions used in neural networks?

The interviewer asked me to explain different activation functions and also asked me to draw their graphs on a Paint App.

I mentioned:

• Sigmoid
• Softmax
• ReLU

I also drew the graphs of these functions on the Paint App.

Question 4: Explain the convolution process.

The interviewer asked me to explain how convolution works and asked me to demonstrate it using an example on a notepad.

I explained the process using a 3×3 kernel and a 5×5 image grid.
I explained that the kernel overlaps with a portion of the image grid.
The corresponding values of the kernel and image are multiplied, and then the resulting values are summed.
The kernel then slides across the image grid, and the same process is repeated from one position to another.

I explained the process step-by-step from the top to the bottom of the image grid.

Question 5: What is the output of the convolution process?

I answered:
Feature grid.

Question 6: Are matrix multiplication and convolution the same?

I answered No.

I explained some of the differences between them.
In matrix multiplication, we perform multiplication between the rows and columns of the matrices.
In convolution, we use a kernel and perform a dot-product-like operation over overlapping regions while sliding the kernel across the image.

I also explained that matrix multiplication can be used for transformation, whereas convolution is used for feature extraction.

Question 7: What else is different between matrix multiplication and convolution?

The interviewer asked for more differences.
After thinking for a while, I said that I was not able to recall any additional differences at that moment.

This was the last question of the interview.

The interviewer ended the round with some suggestions to learn these concepts in more detail and understand their underlying working processes.

There were many questions asked throughout this round, but these are the questions I currently remember.

Round 4: Interview with Founders

The final stage was an interview with the founders.

I did not qualify for this round.

Overall Experience

Overall, the CYNLR interview process was a combination of:

• Mathematics
• Physics
• Programming
• Computer Architecture & Organisation
• Compiler Construction
• System-level concepts
• Machine Learning
• Deep Learning
• Computer Vision

The Online Assessment and interviews tested both technical understanding and the ability to think through problems.

For me, the interview also highlighted the importance of understanding the underlying working process of concepts, rather than only knowing their definitions.

This was my interview experience for the Software Engineer — AI/ML role at Cynlr.