I was motivated to pursue PhD during my master's dissertation work on NLP. I applied for a regular PhD program at IIT Gandhinagar around May 2024. However, I didn’t have any fellowships when I applied. I have been shortlisted for the first round. The interview was completely virtual and conducted via Zoom meeting. My interview was scheduled for 20 April 2024 at 12:30 PM. There were two panelists in the meeting (i.e. panelist 1 was Dr. Krishna Prasad Miyapuram and panelist 2 was Dr. Shanmuganathan Raman). The interview continued for 5 to 10 minutes.
Round 1
panelist1 starts asking questions, as follows
- Tell me about your educational background and research interests.
- How do Transformers work in NLP?
- What is positional embedding? and why we need this.
- In Bengali, do we need positional embedding, and why so?
- What is LSTM and why do we use it?
- What is RNN, and how does it work?
- Do you hard about Machine Translation? Big tech giants like google and Microsoft had already developed machine translation. Why is still research going on it?
- What will be your research objective, if you pursue PhD?
- How do we develop a GPT for Bengali?
- Suvasish, thank you for your time. We should wrap up this conversation. You can leave.
luckily I got shortlisted for the next round of interviews, and they scheduled my round 2 on 22 April at 9:40 AM via Google Meet.
Round 2
I told them, that my research interest lies in the field of NLP, so I was unsure whether I needed to approach any professor before round 2. I went through the faculty profile at IITGN and found there is one professor who is working in the same field which I have my research interest on. His name was Dr. Mayank Singh. I wrote an email telling him, I wanted to work under his supervision, and whether I should mention his name in round 2 if I needed to mention any supervisor's name.
In round 2, there were 3 panelists this time. The first 2 panelists were as before and there was Dr. Mayan Singh as a 3rd panelist. panelist3 started asking me questions as below
- Tell me some of the projects you worked on. (I told them a project based on the seq2seq transformer, which I was working at that time)
- Explain the structure of the transformer.
- How normalization works in the transformer. (not able to answer)
- Do you know self-attention? How it works?
- How word embedding is done in the transformer.
- What kind of word embedding have you done in your project?
- What is your role in the project you just mentioned?
- What is your long-term goal after PhD?
- Panelist 2 asked when your master's will end.
- panelist 1 asked Just to be clear, you have a qualified gate, right?
Conclusion
It was a great experience and hopefully, the dot created will be connected somewhere in my future.