Am I audible?

Yes, Ms., I can hear you clearly. Please start by introducing your professional journey in academia.

I'm now currently working as. Assistant professor at the Apollo University.

Previously I have worked. As assistant professor junior. At VIT University for four years. Before that I have worked as assistant professor in. Seetham's autonomous college.

Right, so you're currently at Apollo University as an Assistant Professor. What's one foundational concept in your field that you especially enjoy teaching to new students, and why?

After finishing by my Btech. I worked in Infosys for three-point.

Okay, you've had a range of teaching roles across different institutions. When you're teaching a theory course versus a lab course, how do you adapt your approach for each environment?

Got it — so you have some industry experience at Infosys before moving into academia. How does your time working in industry influence the way you teach students now?

Yes. Like when it is. Courses. I will. While teaching theory also if it's an algorithm. I want to show the working of the algorithm by coding. Because working in the industry helped me that uh, we should work practically. And, uh, we should. Teach the students how the output changes when you change some parameters like that. So I like to give more. Practical classes, then theory. I'll teach theory, I'll teach the fundamentals, but uh, mostly I concentrate on the. Practical part why because it will be easily understood by the students if we. Teach it properly.

Professor, I noticed your recent publication: "Artificial Flora Optimization: A Novel Feature Selection Approach for Heart Disease Prediction" presented at OTCON 2024. How do you use insights from that research when introducing machine learning or AI topics to your undergraduate students?

Guess very simple. Uh, if it is undergraduate students means uh, they know machine learning now in their curriculum, machine learning is included in. The curriculum of undergraduate students before that also, I'll simply suggest that for example, that artificial floras and nature inspired algorithm where how? Plants will disperse their seeds naturally. Likewise. When you have some. A big task, right? How the according to your problem how you are going to reduce it? According to the propagation distance. The the speed of uh. The plant's speed of raising the seeds or. The when the seats were thrown around that. How it works like that? I will tell them that if you have a larger problem or a problem of larger dimension, what happens is you want to reduce that dimension. For that you can use the some sort of nature inspired algorithm where it optimizes. The way how it disperses it seeds and how the plants will grow naturally. That is that particular algorithm the. Municipal behind that algorithm is. It is used to optimize the number of features in your data set. So. When you run the data set of 1000 features, it takes more time to run. When you reduce those features and you work on it, the time is also less and the performance of your model, whatever the classifier you are using, that will also get increased.

That's a clear way to connect a complex algorithm with something tangible for students. How do you assess whether your students have actually understood concepts like feature selection or dimensionality reduction, beyond just repeating what you've said?

What I will do is. Uh, I will, uh. First of all. Before teaching the. Dimensionality reduction or feature selection itself. I will tell them a natural example. For example if you have 100 bags of rice. And you want a single as a single person, you want to carry 100 bags of rice. That is very difficult. You cannot carry. So if you have more number of dimensions that is very difficult for your model to. Work on it. Right when you have only one bag of rice, what is the difference? So if you have one bag of rice, means when the weight reduces the working speed of. You The Walking speed or the working speed will also increase. Likewise when the number of dimensions. And moreover, you have to remember that the dimension should be. The number of dimensions which you are reducing or removing. Should not affect the performance of your model. Right, it should not affect your performance. Likewise if you are removing some unwanted dimensions or correlated dimensions, then what happens? The speed will of your model or the performance of your classifier will also increase.

Right, so you're using analogies to help students grasp the concept. When you see a student struggling with a practical lab assignment—say, implementing feature selection on a real dataset—what's your approach to guiding them without simply giving away the answer?

Feature selection on a real date as it. First of all, the. I will make the student to understand what are the features in the data set. For example, if the features are. Student ID, Student data set. Right. Student ID, student name, student age. Like that? I will tell him. The unwanted feature is. Name why? Because ID is there one feature is OK. ID and name both means for this one is the ID means he is the person. 2 is the ID means he is the person. It is already noted that ID and name is not going. The name is not going to give. Are going to give uh, any. Uh, uh, it's not going to. Contribute more to the output. So what I will tell him is first you try to.

You try to remove that particular name. Then to age, if you just uh, a data set, which is uh, students age means if the students age, uh, for example, it is uh, some college student is about 18 years or 19 years. Only that is the particular thing you are going to have. I will tell him to 1st clearly note down what are the features that contribute more to his output.

Then I will ask him if you have any algorithm that. That will be directly used, for example in machine learning. I am going to teach in principle component analysis, which is the dimensionality reduction technique. Then I will tell him with the help of principal component analysis. You want to know how many principal components you have. Means how many features you need to reduce. So you first analyze the data set. And first know what are the most important features? What are the most correlated features? If some redundant? Redundant features are there. For example student marks is repeated 2 times, average marks is repeated 2 times. So you can are. Redundant features are there so he can first analyze and study the data set. Remove what is not required and you can keep what is required and you can give the. Output means you can select the important features re and use some dimensionality reduction or feature selection technique to reduce the features and you can get the output.

Okay, that's a pretty concrete way to help them distinguish between essential and non-essential features. Before we move on, is there a particular moment from your research or teaching where a student's question or perspective made you rethink how you present a core concept?

Okay, that's a pretty concrete way to help them distinguish between essential and non-essential features. Before we move on, is there a particular moment from your research or teaching where a student surprised you—maybe challenged your approach or brought a perspective you hadn't considered?

Let's shift our focus to your research productivity and its long-term impact. Given your recent work on nature-inspired AI algorithms for feature selection, where do you see the most promising, fundable research directions over the next three years? Which grants or funding sources would you target to support that agenda?

Yes. Funding. Uh, I haven't, Uh. But. I know many of the people who are doing the funding research means they have applied their applications and they have got it, but I never tried it out. But uh. Uh, for for the next time I should do it? I I want to because people are doing like that. The next three years, I think. The next most of the people would be using nature inspired algorithms or quantum computing. To get the funding. Well of the projects, whatever the projects that are available, we can use this nature inspired algorithms in optimization technique where optimization is required like that.

Let's talk about how you engage students in large classes. Imagine you're teaching a core multimedia or AI theory course with 200 students, but you can't use slides—how would you run a session that keeps everyone actively involved and learning?

Yes, there are many techniques right where we have make them as. Seminars or group discussions. Fishbowl methods. Then Jigsaw methods. There are many techniques to hand large group people. Like we can give. For we can divide the people into groups. Give 11. Topic to study for 10 or 15 minutes and each group will give their answers and the combined answer can be again taught by us and the conclusion can be taught by us like that. And or you can. Make them into. Tell the people to study and come for the seminar. Previously you can inform them if there are large number of students. Instead of slides you can make it something interactive and so other students can also be benefited from that.

Accreditation cycles can be pretty demanding — say you join our department and notice that outcome assessment data isn't consistent across courses. How would you approach fixing that, both in terms of process and getting faculty buy-in?

Or you can have some practical sessions of showing. Some coding or the output or some videos? Or how the images are being done like that.

Professor, let's consider a scenario that comes up in academia. Imagine a student files a formal complaint, alleging your grading is biased. At the same time, your department head is pressuring you to improve pass rates in your course. How do you handle this situation, balancing fairness with institutional expectations?

Biased.

That's just the allegation—I'm curious about your process. If your department head disagrees with how you handle the grading dispute and insists you adjust marks to raise the pass rate, what would you actually do?

That I cannot do. Why? Because then how should I? I can show the difference between the student who studied and who did not study.

So I will. I will clearly explain the situation to my department head. Who is ever pressuring me like? This is the situation. I cannot do that.

Let's shift to how you bridge academia and industry, Professor. Given your experience at Infosys and your AI research, do you have industry connections that could help translate into internships or placement opportunities for your students?

And I will try to make the students from the next time. I will try to make the students from the beginning itself to study and to. Get good marks in the exam, that's what I can do.

Let’s shift gears to a few targeted skill areas required for this role. When serving as an examiner—say, for a final practical exam in an AI or multimedia lab—how do you ensure your grading is consistent and objective across a large group of students?

Okay, that's a firm stance. Let's move to your doctoral work. Professor, what was the central research question of your PhD, and how did you ensure your findings were both original and significant in your field?

Yes, my topic was that. Bio inspired optimization frameworks for. Heart disease prediction.

Right, so your PhD focused on bio-inspired optimization for heart disease prediction. What concrete steps did you take to validate that your framework actually improved prediction accuracy compared to existing methods?

With the help of the other people and my guide I tried to work on my code and I get the result. I got the results and that results I have published in my papers.

Right, so you iterated on your code and compared results with published work, getting input from your advisor and peers. What did you do to ensure your methodology couldn't just overfit to the specific datasets you used—did you validate on external data, or use any particular statistical tests?

Yes. Yes, uh, in my research papers itself initially when I uh, gave the results like that, I got the same question that how you do the external validation. You can use the other data set and validate it. I have done it. For heart disease they have asked me to do the same methodology on PYMA diabetes data set. I have done that also on some statistical tests also. They told me to do statistical T test, Wilcoxon test. I have done those testings. And I have. I have got the output. Whatever I got the output, I have made it as a table saying that significantly. Significant improvement is there. In for one data set I got significant improvement, the other data set I did not get. But anyway that I have kept in my results saying to the saying to the reviewer that when the reviewer asked these questions, saying to the reviewer that this is the output I got if something was to be done. If you let me know I will do it. That is what I have done.

You gave a clear example of validating your approach across different datasets and using statistical tests—sounds like you got some pushback from reviewers and addressed it directly. Let's talk about your industry connections for consultancy or projects. Have you ever worked with a company or external client to apply your AI or multimedia research in a practical setting?

No, no, I haven't done up. Approach any company or any other client. Uh, why? Because uh, till now I did not get any chance of doing that. If in future if I get a chance I will try to do that. But one thing I wanted to do is with the help of the students whom I'm teaching, I want to build some model where that model should reach and help the doctors to do at least it should reduce some. Umm, part of the. Work of the doctors I'm trying on it. Working on it. We'll see.

Alright, since you haven't worked directly with industry or consultancy projects yet, what steps would you take to start building those connections—especially if you wanted to involve your students in solving real multimedia or AI problems for external partners?

I will, uh. Ask my students. Uh uh, to work on what are the real time problems that are. That the people are facing in AI or in the consultancy or in the project. VE with the real we will first. Uh, take the problem. And we will outline what solution we can do and we will approach that particular consultancy over the company and we will tell them for your, uh, what are the problems existing, uh, with your, uh. Consultancy or industry, this is a solution we are trying, we are having. If you are having any plan to work with us, we will. Coordinate with with you and we will do the. We will give the solution for that particular work. And we will tell the details that are asked by the company. Like what are all the needs and requirements of the company that we can do and give them?

Alright, we're wrapping up. Is there anything you'd like to ask about the role or the team?

Boatroll.

For this position, the main responsibilities are teaching and mentoring students in emerging tech specializations, guiding projects, supporting research, and contributing to both publications and academic activities. The department values adaptability, strong subject expertise, and a real commitment to teaching excellence. Anything else you'd like to know?

But I like to know for. Vit you are taking this interview or not?

That’s a fair question. I’m conducting this interview as part of the academic search process, but I can’t comment on specific institutional details. Is there anything else you’d like to ask about the position itself?

Nothing. Thank you.