Yes, Sir, we can continue.

Please start by introducing your professional journey in academia

Uh, good evening, Sir. I'm Susie. And then I had. Uh, completed my. PhD erata correction. I'm waiting for my Viber's. And then I did my master degree. E-mail Anna University, Trichurapalli and then. My UG. In under Madras University. And then in 2004 I completed my UG degree. Then I'm having nearly 13 plus years of teaching experience. When my interested subject is. Uh, data structure. Operating system. And then deep learning. And then my year. Master. I mean like my PhD research work is. Based on social network analysis like I'm analyzing. Misinformation detection. Uh, considering multimodal messages. And then identifying influential user.

I would like to work. On research. Area. As well as guide. Students in academic as well as. In projects. And then I would like to. Jointly work. And contribute to the benefit of our institution.

So Mrs., you mentioned you're passionate about teaching data structures, operating systems, and deep learning. When you introduce data structures to a new class, what's the first concept or example you use to get students interested?

Thank you.

Okay, so you see yourself guiding students in both academics and research, and collaborating on institutional projects. When you work with students on their projects, how do you make sure they actually understand the underlying theory and aren't just implementing code or following instructions?

Alright, Mrs., I hear your interest in guiding students and contributing to the institution. Earlier, you mentioned your passion for data structures and operating systems. When you teach data structures, what's one example or analogy you use to help students grasp a foundational concept, especially for those struggling with abstract ideas?

OK, here data structure is that for example. We we need to organize the data whatever we are handling. So in that case. Since we are computer science background. All the datas are going to be represented and stored. Whenever it is stored only we are able to access it. So what are all the structures that is existing? In order to handle the data I will be. Teaching with an application. Oriented to the students and then I'll be guiding them. Step by step as it is related with the programming concept also. So I'll be explaining them the program.

I'll be explaining them, the program and the step by step. Uh, execution.

Okay, so when you teach data structures, you focus on how to organize and store data, and you tie it to practical programming examples. Walk me through a moment when a student was really struggling with a core concept — how did you help them understand it?

So that they will know. Come to know the real world applications as well as. How the datas are organized and accessed and managed in the computer system?

You focus on explaining programs step by step, making the process clear with examples. Can you recall a time when a student was still confused after your initial explanation—what specifically did you do to help them understand?

Okay, so you emphasize step-by-step execution and real-world context when teaching data structures. In your experience, what’s one practical application you use to help students see why linked lists or stacks actually matter outside the classroom?

Like for example if we uh go for a stack applications. So they are used in like programming concept of recursion. As recursionist. Invoking the function. Uh, within itself. So in that case. Uh, this is, uh, stack. Follows the operation of last and first out. So last in, first out means whatever the. Data entered last or whatever the data entered the. Uh, or process or job entered at the last question will be. Uh, popped out for the. For the further execution likewise. Uh, the stack applications. For example in Word processing we can consider. While you're editing your document. By unknowingly if you had deleted your. Data so through your undo. You can regain your operation. In that case the last and first out operation is up. Apply so this I can give as an real time example for the students to better understand how the stack. Stack operation works out. And then considering linked list. Now for example stack is going to be here. Linear way of storing data. In that linear way of storing data, that is a memory wastage. So in that case. Uh, we need to go for. Nonlinear data structure. In non-linear date. Uh, nonlinear. No. OK, LinkedIn list is also a linear data structure, but not in a, I'm sorry. I'm little bit confused. Here I want to make it clear that. Stack as well as linked list is going to be a linear data structure. So in that case. Like early what is the major difference is going to be is in stack. You need to allocate the memory in a continuous way. Whereas in LinkedIn. The memory is allocated. Whenever necessary or whenever data. Exist, so a pointer is. Used for the connection. In that case. Where we will be using this linked list concept is for your memory management concept. Where do you want to store the data? Not in a continuous week. So this is what I would like to explain the students.

Mrs., I see you published “Utilizing rPPG Signal Synchronization and Deep Learning Techniques for Deepfake Video Detection” in IEEE Access. How do you bring insights from that research into your teaching, especially when explaining deep learning or artificial intelligence to students who may not have much background in the field?

OK, it is related with your. One of the. A part of fake news detection. One of the modality I had taken as fake video. Like in our social network, we users are, uh, connected to the. Network social media. Uh, like, uh. Nearly for three hours or 4 hours, we spend our time in online. So in that case. When when we encounter. A message. We are not knowing whether it is a true message or. False message. In that case we we are in a need to identify. The deepfake video one of the. Modality is I had taken as a video, so deepfake video had considered. In that case. Now how the deep learning concept is implemented here? For the fake video detection is that. Ask uh deep learning concept. Involves training. More data and then. Finally getting out the final result. So in order to train more data here. The videos, real videos and fake videos are. Given to the proposed model. And then it is processed and finally we'll be obtaining the classification of. Real love fake video. So the novelty here is that. Normally a video contains. We'll be in, uh, we'll be capturing the face. Of that, uh, we like, uh, each video is. Obtained as a frame by frame for analysis. And in each frame. We'll be capturing the face. And particularly the face. As my novelty is focused on the synchronization of the signals between the. Two regions. Here the two region is choosed as the. Left and the right cheek in that case. I need to go for the synchronize synchronization of the signal. So here. Uh, since it is dealing with remote photoplethismography signal. It is uh, like, uh. Identifying the. Colour which from the cheek regions. So in that case. Uh, we are trying to obtain the metrics of that signal. And then trying to find synchronization between the left and the right cheek. And finally, we are. Incorporating the spatial and temporal. Temporal data. So that. We are identifying by using these things. We are identifying whether the. Video is real or fake. So as it is one of the main project for the societal contribution. As many youngsters nowadays are for. Spending their much time in the. Internet. So we are in a need. To identify whether the data is. Real or fake? Thank you, Sir.

So your deepfake video research hinges on analyzing synchronization of facial signals and leveraging deep learning for classification—interesting way to connect technical novelty to societal impact. When you teach undergraduates about deep learning using this project, what challenges do you face in making the signal processing and model training accessible, and how do you adapt your explanation for those who have never worked with neural networks before?

I can relate the neural network concept with our brain. How our brain thinks and how we relate. Each incidents, whatever happening. And then we are trying to identify the pattern among the events happening. And finally, we will come to a conclusion when certain events taking place at that particular time. This is how the neural network also keeps on working. When we given input to the. Neural networks. So they waits and bias is added. And then? Uh, each neuron. Tries to learn. The input. Based on the dimensionality. Whatever the datas are given. And after learning it will be passing the. The output to the. Next to neurons. Likewise layer by layer. It passes and finally when it reaches the classification layer. The output is predicted. So here we will be comparing the. Predicted output with the original output. And when there is a mismatch when it is not meeting the threshold point. So error rate error. Calculated so the error will be. Like, uh. Accurate minus predicted. So whatever the. Balance amount. Balance value. Is calculated that is going to act as a. Umm lost, so in that case. Our deep learning. Mainly Howard. Trains and learns the. Patterns is that. It will try to minimize the loss. So that. Accuracy. And predicted value. Merely meets the. A greater threshold value, so in that case again. Back forwarding will be. Uh, happening like. In each. How much loss had? Taken place which? Part of the. Data or which part of the weights and bias? Is to be adjusted. So that. The. Laws can be minimized. Likewise this. Keeps on uh, training. Each each time, each iterations. And then it it. Tries to learn the proper weight and bias. For each. Dimensions and then finally. The output will be predicted so this doesn't happen at one time. It keeps on learning. The deep neural network keeps on learning as. We also learn in our everyday life whatever the events happening. And we store it in our mind. And finally. When another event. Happens. We try to relate. The happenings, what we faced in our life and then finally we will be concluding that. So this happened because of it. Likewise. The deep learning network also, it is all about training. How the data is? How the weights in Bhayas are? Uh, reassigned. Based on the error. Faced and then finally the outcome will be. The given.

Alright, so you use analogies to the human brain and learning from experience to make neural networks relatable for students who are new to the concept. Shifting gears a little, you mentioned guiding students on both academic and project work. Can you share a moment when you helped a student bridge the gap between theory and application—maybe in a project or lab setting?

Thank you, Sir.

Given your research on multimodal misinformation detection and deepfake video analysis, where do you see the most promising—and fundable—research directions over the next three years? What kinds of grants or external funding sources would you target to support this work?

You mentioned earlier that you focus on application-oriented teaching and guiding students through programming concepts step by step. Suppose you’re teaching a large introductory course—say, 200 students—and slides aren’t allowed. How would you keep everyone engaged and actually learning the material during a session?

OK if. Earlier we were following only more teaching method. As smart classrooms came only, we are utilizing the projector and the slideshow presentation. If slideshow is not allowed. Then I'll be using the. Teaching method of using board, chalk and top. So there I'll be writing the program. And then I will be explaining them. Line billing because. My one of the professor explained me the concept of. The CC plus plus programming by writing the programs on the board and he started to explain 9 by. That created more impact in me. And I also follow the same kind of pattern teaching in my lecture hour also. So by that way, students also will keep more attention. How the variables values assigned and how the functions are called? So I can. Uh, like interactively? I can explain them line by line. How it is executed? I can give them an imagination of perception there there.

Alright, so you’d use the board and walk students through code line by line, which matches how you learned C and C++ yourself. Let’s say some students zone out or aren’t engaging, especially in a class that size—how do you spot that, and what would you do to pull them back in?

Uh, like I'll be, uh. Pointing out them. And asking questions. Uh, from the lecture what I had delivered. When they are not able to answer to the questions. I'll ask them to listen the class. And then? Kindly attentive to the classroom. Uh, like that way I'll be asking them as well as. Uh, like if that person is not answering. I will post that question to all other students so that who are all not active in the class, they might also get. Interactive in the class. When we when there is. Going to be a two way communication between the student and the staff. Then obviously the student will not lose the interest. I think so.

Thank you, Sir.

Okay, so you rely on direct questioning and spreading questions among students to keep a large class engaged, especially if some aren’t paying attention. Let’s shift gears—accreditation cycles can be demanding. If you joined our department and found that outcome assessment data was inconsistent across courses, how would you approach fixing that?

Sir, I'm not clear with your question, Sir.

No problem, Mrs. What I mean is, sometimes the way we collect or report student learning outcomes varies from one course to another, which can cause issues for accreditation reviews. If you noticed that kind of inconsistency in your department, what practical steps would you take to make sure everyone’s assessment data aligns and meets university standards?

If. Uh, inconsistency is. Met in the department then it is our departmental issue as well as staff. Who is in charge of it? Should need to take the responsibility of it. To correct the. Uh, data data and give the correct data reported data. So we need to sort out Sir.

You you can take help of another staff.

Or a friend and then sort out where the mistake had gone wrong. And then finally. Uh, reorganize and give the corrected report to the department, Sir.

Okay, so you'd see it as a shared responsibility for the department and the staff in charge to correct and align the reported data. What practical steps would you personally take—would you call a meeting, create a checklist, standardize the process, or something else?

Okay, so you'd see it as a shared responsibility and involve other staff to resolve the inconsistency. But practically—would you call a meeting, create a shared template, or do something else to make sure everyone’s outcome data lines up across courses?

Alright, so you'd work with colleagues to pinpoint where the data went wrong and get it corrected before submitting. How familiar are you with accreditation processes themselves, and have you contributed to institutional quality assurance or accreditation work in your previous academic roles?

OK, I had worked out for uh. NBA aggregation. As well as uh. Uh, the other college datas are maintaining. We'll be assigned as a mentor for one class. Suppose I'm assigned as a mentor for a class. Then. I will be uh. Holding all the student profile information, I'll be responsible for it. As well as. Uh, if the students are participating in the. Events or any cultural programs or sports events or any hackathon programs. I'll be collecting the datas and. Keeping it whenever it is asked for aggregation purpose. When I'll be submitting those reports to the. Faculty in charge. Poor assigned for it, Sir. I had supported, I had worked out for NBA aggregation.

Okay, so you’ve handled NBA accreditation duties before—collecting student activity data, acting as class mentor, and submitting reports to faculty in charge. When you’re evaluating students from different backgrounds or abilities at VIT, how do you make sure your assessments are fair and accurate?

We can ensure the fairness as well as the accurate. Uh, accurate? Accurate. Accuracy is. Uh, like through inquiry only, Sir? Like, uh, we need to collect. If the data doesn't. Belong to our department or we are collecting it from another department? It is through inquiry. If the student is present there, we can call them and uh. Finalise whether it is correct or not. Otherwise we need to make a call and then ensure. The date as well as the event whether he had attended is correct. So through inquiry only, I. I will be able to give the correct.

Let's consider an ethical dilemma you might face as a faculty member. Imagine a student files a formal complaint alleging your grading is biased, but at the same time, your department head is pressuring you to improve pass rates in your course. How would you navigate that situation?

So the fairness of the outcome is based on the continuous assess. Like we'll be handling them. Uh, module test. Or unit test will be conducted. As well as any projects will be assigned. As well as attendance will be considered. And then, uh, finally.

Any assignments will be. Assigned. The final continuous assessment mark only will be given as an. Internal mark to the. Uh, department. So there we cannot compromise. No Sir. It is based on the performance of the student. We need to give the. True information unless otherwise. He's facing any mental. Uh, mental disability? Or. Any accidental uh. Medical related problems he had. Then we need to contact him retest once if he had recovered. We need to conduct him a retest. Or else if the failed students in the continuous assessment need to be motivated. With the help of the department permission. Uh, with the guidance of the department. Then I'll be conducting them. A retest. And then I will be boosting them. So through that way we can compromise. The this all can be done only with the approval of the department. Not by the individual member.

Alright, so you use continuous assessment methods like module tests, projects, and attendance to keep things fair, and you mentioned verifying event participation through direct inquiry with students or other sources. Let’s talk about something trickier: Imagine a student files a formal complaint saying your grading was biased, but at the same time, your department head is pushing you to raise pass rates. How would you balance upholding academic integrity with that kind of institutional pressure?

Alright, so you make sure fairness comes from continuous assessment, verifying participation directly, and only making exceptions for serious issues, all with department approval. Imagine this—one of your students files a formal complaint that your grading was biased, but at the same time, your department head is urging you to improve pass rates. How would you handle that situation?

So he cannot claim me as a biased because I'm having the final report of the continuous assessment whatever I had conducted. So I will be giving them the grade based on the. Mark only. So in that case he cannot claim me as I'm a biased. Staff unless otherwise he have a proof for it. And then when my head of the department is urging me to pass. On what? A fact the head of the department is. Asking me to. Put past percentage rate. That need to be discussed with the department Head, Sir. So if it is going to be. Uh, like, uh, acceptable. Then I need to follow the head of the department order. So in that case. With the guidance, I'll be doing it.

Alright. Given your work in both teaching and accreditation, let’s talk about bridging academia and industry. In your experience with multimedia or artificial intelligence in media, do you have any connections with industry or professional organizations that could help students access internships or placements?

Yeah, Sir. Uh, like. Students need to undergo internship. Then only there'll be no. Knowing how the theory as well as the practical. Goes in hand in hand to hand. Or it is of indifferent way whatever the learning is going in a theoretical way is. Basic and what they are implementing is the real knowledge they are gaining. So in that way we need to have. A tie up with the industry as well as the academic institution. To have a balanced learning. As well as, uh, implementing. So I'll be encouraging the students. To go for the industrial experience. As well as. Uh, even teachers also need to have that experience. Hence as it will be helpful for them, we are have We are undergoing faculty development program. So that also is one of the factor we will be learning every day. And along with that knowledge. Even a practical knowledge will be. Helpful for the teachers as well as the students. To undergrow. A combination of projects. As well as I would also like to. A request for the. Inter departmental activities so that. A combined. Project can be. Uh, beneficial for the societal? Uh, contribution. Majorly for the agricultural industry. So to promote the agricultural work. It is better to go for the. Inter departmental project, so I'll be. My aim is to. Uh, focus on that area and then I will be requesting for this. Uh, idea to be implemented? Thank you, Sir.

Okay, so you see value in building those industry and cross-departmental collaborations, especially for real-world impact like agriculture. Shifting gears, Professor, I’d like to quickly check your experience in a few key areas for this role. Have you directly worked on multimedia or artificial intelligence projects applied to media, and if so, what was your most hands-on technical contribution?

OK, my technical contribution is that. I. Two models. One is I, uh. My proposed novelty work is that I evaluated the. Influential user. Score like uh influence driven diffusion score. What that score like? I'm trying to incorporate user engagement metrics in that. Because. Not only the centrality measures helps to identify the influential user. But. The. Uh, engagement metrics also plays a major role in identifying the influential user likewise. My another contribution of novelty is that. Uh, utilizing the synchronization of signals. RPPG signals. To identify the. Fake video. So. While incorrect while implementing this uh. Multimedia project. I utilized deep. I used deep learning models. So the. They I incorporated. Uh, like a neural network? Models and then. Trained. For a long time and then finally I obtained the output. So my proposed model achieved. A good accuracy rate of 94.6 percentage.

Thank you.

You mentioned two technical contributions: developing an influence-driven diffusion score that incorporates user engagement metrics, and building a deep learning model for fake video detection using rPPG signals, achieving over 94 percent accuracy. That’s quite specialized work. Alright, we're wrapping up. Is there anything you'd like to ask about the role or the team?

Yeah, I don't understand your question.

Right, Professor, thanks for clarifying. We’re just about wrapped up. Is there anything you’d like to ask about the role or the team before we finish?