Uh, is speaking.

Alright, just to get us started — could you walk me through your professional journey in academia?

Yeah, sure. Uh, good evening myself, Doctor Jennifer. I have completed my PhD in the year 20. 22. From Anna University. Umm. And uh. I have completed my. PG in 2009 from Anna University, Trichurapalli, 23. The 2007 I have completed my. B in computer Science and Engineering from Jayarajanapakam CSI Engineering College after completing my. PG By 2009 I joined in a push link engineering college. Work there for four years and then move to RMD Engineering College, Chennai. And then I moved to KPI instead of engineering technology and right now I'm working in Madanapalle Institute of Technology and Science. Throughout my 16 plus years of teaching career, I've completed around 17 research papers that includes conference, Scopus, indexed conference publications. And then I have completed 3 consultancy boards with the next 99 from Chennai company and then. Uh, standard insights, Chennai. I have completed these, uh with these two companies. I have completed 3. Consultancy works worth 2.5 lakhs. And moreover. Through the company interaction, the industry interaction. I, uh, got the opportunity, internship opportunities for. More than two students and then. I used to handle the subjects like deep learning data structures. C programming. Python programming. These are all my area of. Request and in my research also I was mainly concentrating on image processing using deep learning for various different medical and other applications. Thank you.

You mentioned handling subjects like deep learning, data structures, and C programming. Out of all the courses you've taught, is there one foundational concept you especially enjoy introducing to students?

Uh, yes, when it comes with the subject. Deep learning I really enjoyed when. I thought the concept of neural network to students because. Nowadays, wherever, whatever the applications models we are concentrating on, everything is based on. A, A and whereas image processing. When it comes with I would like to. Introduce my students about the concept of neural networks and how. Especially the CNN model can be used for. Image processing, how the features can be extracted so in that. The case I really enjoyed in that and then when it comes with the programming concepts. Like Python And C programming. It is really good. I felt good that I'm teaching the students the. Fundamental concepts of how actually to think logically. And how to solve those logical? Into a. Problem using uh C? Or Python program. So these two subjects I really enjoyed when I was teaching to my students.

Shall we move to the next?

Alright, so you enjoy teaching neural networks—especially convolutional networks for image processing—and you also like helping students build logical thinking skills in C and Python. When you introduce something like CNNs to students for the first time, what's your approach for making a challenging topic accessible to those who might be intimidated by the math or complexity?

So when you introduce neural networks, especially CNNs, to students—what's your approach to making those abstract concepts accessible for someone who's never seen them before?

When I was started the neural network concept, I started with a normal human thinking. How the neurons in a human brain. Things when I view. See an image how actually as a. Kid. The kid starts to learn about the image, whether it is a green apple. Or red apple? How it is actually identifying whether it is Apple? Or orange using the features so the neurons in human brain make them to understand. Buy day by day. By learning. By teaching by the. By the parents, by some elders or by themselves by looking at the image. How the kids they are neurons are learning the brain neurons in the same way. I tried to teach my students. Like how the brain thinks? Uh, first time when I see an apple. I don't know but my mom thought me. This is an apple, so my new run started to learn. In the same way, the neural networks also started to learn. From. The models that testing images, what we are processing. And that is how why? The importance of data plays a role how the image data should be, whether it should be labeled as clear data, the importance of preprocessing So with this I started and they thought those subjects that. Concepts to my students.

Okay, so you ground abstract neural network concepts in how the human brain learns, making it relatable for students, and then connect that to why labeled data and preprocessing matter. I noticed your 2025 publication on “Cervical cancer prediction using deformable kernel darknet-53 and depthwise separable convolutional neural networks” in Scientific Reports. How do you bring insights from that research into your undergraduate teaching, especially for students new to medical image processing or advanced neural architectures?

So from that particular paper. I can say like. 1st I have to make them to understand the medical imaging. Because. When I say a normal image data set like fruits. But or. Animals. That is different from the medical data set. So first I have to. Make the students because they are very new to the medical data set medical image processing. So first I have to make them to understand what are the. Difficulties we may face with. Medical images, for example, the noise present in the data. The importance of identifying the disease or the infection. Predicting the infection accurately then. The normal data set. So first I have to make I will make the students to understand the difference between a normal medical day, normal data. Image data and a medical image data. With that I can make them to understand first the data clearly. And then I can make them to understand. How the features can be extracted? For example here I used a document so insteadof.net OK. Can you CNN if I'm using CNN resnet or some other? Architecture. What kind of difference I mean? Peace with respect to the feature extraction. Likewise, I can teach the insights about these papers to my students.

That's an interesting point. Could you elaborate a bit more on that?

For example. Uh, I'm taking an image, an MRI image. 1st I have to remove the noises. So I have to teach the students about the preprocessing method. Uh, when I'm teaching the preprocessing. There are so many methods available, so. Can I tell them? All everything at the time, no. So I will start with anyone of the pre processing technique. For example I'm converting my image into an excel where some. Uh, invalid data are there are. Uh, fields are there which are supposed to be removed so I can. Tell them how. Uh, noises are removed from those image data set or in any general data set. So first I can teach them that about the preprocessing. So. Once it is over. The students may get the idea or interest about how actually the data is preprocessed now it is ready for processing and then I can teach them how CNN is extracting the features. For example, the features of all the bones is not required when our brain image is not required or when I am coming with any. With the fungus images like umm. I'm coming with the. Bane identification not all required Only if there is. Swelling in the vein so by that time I can make the students to understand these are all the important features. So from that they are supposed to extract the. Those values.

Right, you emphasize first demystifying medical image data—teaching preprocessing step by step, then focusing on extracting relevant features for things like vein swelling, so students see the reasoning behind each stage. Shifting to research sustainability for a moment: given your expertise in deep learning for medical imaging, where do you see the most promising, fundable research directions over the next few years, and which grant sources or funding agencies would you target?

Yes, I would like to, uh, concentrate on. DSD. Arts uh, recently. As I'm right right now I'm working in Andhra Pradesh Madanapalle Institute of Technology and Science. Uh, with one horticulture university through college, uh, I developed one proposal. It is still under. Consideration not. Approval that. We tried to provide. Disease identification mainly for animals. Because being a human, I can say. Well, uh, I got the. Stomach pain. Headache. Likewise, I can clearly say when it comes with animals it is. Difficult to identify, so bovine uh. Managing, managing the health of bovine, that means cattle. So for that I prepared one proposal. So I am. Planning so far with Surve. Then DST. I'm planning. To provide with respect to medical not only for humans but also for animals. How? This kind of image medical image processing can be extended with that I'm planning to. Prepare a proposal. Within the next one year after joining the vid. And moreover, the budget would be around 20,00,000 and the duration of. Two years. This is what my plan so far. Maybe within these time period if I come across any other? Methods. Our technologies are challenges. Posted by anyone of the DST observed kind of platform. Then I may also choose the and work on it to prepare a proposal according.

So you've already put together a proposal for disease identification in cattle, targeting DST and focusing on animal health—twenty lakh budget, two years, and some industry collaboration. What would you say is the biggest risk to sustaining this research as a long-term institutional asset, especially if the initial DST funding doesn't come through?

I'm planning like not only in DST. Uh, maybe I can even. If uh, I'm not having the possibility of getting selected in DST. Uh, there are some funding. Agencies they are. Throughout the year. They are giving the opportunity for provide for preparing the project proposal, so I think. Even I can make use of them. To do it. Um, and other than that. Uh, there are certain industries where there is no. R&D separate R&D department because of. Because that industries are startups, so. I'm planning to have an interaction with the industry. So that not only. Through anyone of the government professional bodies, but also through. Industries if it is possible to generate the revenue by doing the consultancy work. Are any R&D work for that particular startup industries that is also in my pipeline? So that what I can do is. By the time of preparing for DST and other kind of. Platforms this also side-by-side I can do. So that I don't need to. Rely only for the DST proposal. And it's funding. I can even use the funding. The external private, uh. Uh, industries. Startup level industries especially. Already I'm in contact with the 2 industries next to 99. And standard insights from Chennai. Through which I I'm in contact with their chief technical officer. So I think it is quite possible for me to communicate with them to get the funding for my. Anyone of my research one?

Given your experience teaching deep learning and image processing, imagine you’re running a session for 150 undergraduates—but you can’t use slides or a blackboard, just yourself and the room. How would you make that session engaging and ensure students actually grasp the key concepts?

Uh, as I'm using the slides, uh. PowerPoint slides. And if there are 150 students, undergraduate students, those are very new. To the concept of. Deep learning or neural networks. OK So what I can do is first. First of all. I can create people, I can create a slide which contains as I mentioned, what is a new run. What is the importance of neuron for a human? And then? How that neurons can be converted into a neurons of a deep learning model? So first let me teach them just what is a neuron, how actually it can be represented as a picture and how the weights are passed, what is the value that I'm giving for that neuron. So this one after teaching this. I can make the students. Calculate the weight the output passed from. One neuron 2. Other neurons. Using the formula W star WI Sigma is equal to. UH-12NA WIXI plus the bias value. I can teach them. To calculate it manually. So that by giving. This kind of hands on problems. The students can be. Ever Like how actually theoretically? With respect to mathematical calculation, how the weights outputs are calculated and how it is getting passed from one neuron to other neuron that the students can able to understand. So by this this is one way of engaging the students with respect to those key concepts and then if it is possible with respect to. Their collab Google collab platform I can make them to work. With the Python programming, how actually the neuron is getting? Created. Or or at least virtually, how the values are calculated and passed to? The other umm. Neurons that subsequent layers of neurons. So this is the basic level. So then I am teaching coming to the concepts like for example. Activation functions. So by that time I can give them the problems. That means case studies like when the value of. Baits are updating. Of actually the next two data, next layer of my neural network is going to be affected. When applying the activation function, for example the softmax OK, how? The results are going to be at the last final output layer. How it is going to get the OR select the? Output predict the output. So these kind of things with respect to. Mathematical calculations. Plus if possible. With the Google collab environment I can make them to work. So that. It is. Easy to manage all the 150 students as well as it is easy for the students to understand. From the basic how actually mathematically, theoretically, and then with respect to program, how the neurons are getting? Implemented and how they are passing the information. All these kind of data can be understandable by them.

Right, so let's talk a bit about responsibilities beyond teaching and research. Accreditation cycles can be demanding—say you join our department and discover that outcome assessment data is inconsistent across courses. How would you approach fixing that?

So. If the outcome assessment data is mismatching. I can communicate with the Coe. Department level Examination coordinator are at the. A criterion coordinator are. The overall institute level. Uh, coordinator. Uh, for regulating the assessment methodology, to the best of my knowledge, wherever there comes some lagging or there comes some. Mismatch moreover, when it comes with outcome based education if I'm concentrating. During the development of my curriculum itself, I hope. That would reduce the problem of these kind of assessment mismatches at the end of the. Academic year or the semester? So. Uh, from my side, I would like to contribute to the preparation of the syllabus itself. Based on? Purely based on. Outcome based. So for example. I cannot start any subject directly from the analyze level. At the level unit 1 I can start from the understanding level, so. Uh, I I will be helpful for. My higher authorities for my VO is NBA coordinator or my head of the department to frame the syllabus itself how it can be aligned with the outcome properly so. If this is. One point completed when we are preparing the questions for the assessment. There comes how I'm matching the outcome with a question. I should not exit the level of my outcome. Uh, unit 3 is from the level applied as per Blooms take tsunami level it is applied, but I'm asking a question that from analyze or create our design level that is inappropriate. So during the preparation of the question paper itself in alignment with the. Course syllabus. I can enable the. Faculty are the team to prepare the question accordingly so that. During the assessment, the mismatch will be. Heavily it will be reduced. Or almost no mismatches will happen unless otherwise any manual problems occur.

Professor, imagine a student alleges that your grading is biased and files a formal complaint, while at the same time your department head is urging you to improve pass rates across your courses. How would you navigate that situation and uphold both academic integrity and your obligations to the institution?

So now first of all, being a teacher for 16 plus years. I would like to avoid that bias. Uh. Marking system. One day purely based on their performance, but still if the student is complaining. About my bias, the first thing is always better to have a pakka back back up like a report whenever I'm giving the marks. Based on what I have reduced the marks. Based on what? Which error or? The mistake of that students answer. I'm reducing the mark. That kind of. Food can be done completely. That is one kind of. Uh, one kind of maintaining that. Balanced assessment method. Another one is if my. Umm had this. Asking me to. Improve that. Past percentage, The thing is. But to maintain the integrity as an academician. I can make the students. Those are already very low in performance, I can call them. Personally and I can ask them like what is the difficulty they are facing? Maybe. If the student is having any. Personal problems I may be. Uh, asking them to prepare next time I can give them more chance, more time. Personal cat to sit and. Prepare for their their assessment. Or if the student is feeling like technically they are not able to cope up. Then I can make them to sit and. With them I can sit and I can. Do either the peer learning that means the students those are doing good with them. I can make a. Thing and I can make them to prepare well or as an individual I can sit with them and I can make them to understand the concept so that they can perform well in the exam. So being a teacher, it is my duty, my responsibility to spend more. Time with each and every individual student to improve their performance so that I believe. This will maintain my academic integrity. Without spoiling the students learning culture. I can make them to get. Pass as well as I can make them to understand the concepts very clearly because. Understanding the concept. Is the major thing when we are learning any subject. Not just getting past so I will concentrate on making them to end the step. Through that I will try to make them to. Paths so that I can both satisfy the students requirement as well as my HEAD requirement.

Okay, so you've got a habit of documenting exactly why marks are awarded or deducted and you're proactive about one-on-one support for struggling students—that's clear. Let's shift to something VIT values a lot: industry alignment. Given your background in deep learning and image processing, do you have industry connections or collaborations that could help students secure internships or bring real-world projects into your classroom?

Uh yes, I have a collaboration with the 2 industries Next 99 and standard in size private limited. Both are from Chennai. So through that. Sorry. Through that for the past. Uh, two years left. I made three students. To get internship with this type and. And from the students feedback, I came to know that the students in a live. Real world environment from the industry they can able to learn. The practical problems, the approaches they are supposed to follow in an industry, They learned it very clearly. So it is. Completely possible. In the same way, I can make an industry. To have any interaction with my students either the offline move in offline mode. Or in online mode I can arrange. That kind of interaction with industries. There are people because I have a connection with many of my old students, those who are right now working in industries, so it is possible. Either through. The direct industry presents itself. Or by having the internship. In any industry. It is possible.

Alright, Professor, I want to ask about your experience with student evaluation and exam duties. Say you’re responsible for coordinating end-semester exams for a large class. How would you ensure fairness and consistency in setting question papers, invigilation, and grading?

When it comes with requesting paper. I have to align with the syllabus. Whatever the cause outcomes mentioned in the syllabus, that is why I mentioned. Preparing the curriculum itself. Plays an important role when I want to inculcate an integrity in academia and I can make the students to. Fairly make them to learn and get. Pass so based on that outcomes and the subject syllabus I can prepare existing paper. Whereas it is better always to prepare a question paper. In such a way that I should not ask all the questions at the analy and apply level. Because there may be some students who are in the average level, so for them to. For them also to get pass in these kind of examinations at least some 40 to 50% question I can ask in understand to. Remember, understand and then apply level. Remaining 50 students were very capable. Those who are creamy layers students for them I can ask the questions in apply and and analyze level, if possible in the design level in case of laboratory examinations. During as I'm setting the question in this way, it is easy for both the average level students get passed as well As for the. Creamy layer students to get. Good marks. So this is with respect to the question paper setting. As the invigilator, whatever the instructions given from the Coe of the institution, I will follow as it is, for example, keeping no mobile phones, no cheats. No other kind of cheating method, so I used to check the respect to the Arab. Identity card with their hall tickets and then only I can let them inside. During my invigilation duty of whether it is 1.5 hours or three hours, I'm supposed to be very vigilant so that the students are not able to pass the cheats or do any kind of malpractice though that I will do as an invigilator. As an evaluator when I'm trying to assess the question paper. Better to create always. Key based on the questions given. So sometimes the students may answer exactly the definition as it is in the book. But some students are not like that. They may. Given the context of the answer is correct, but not the exact word. So being in evaluator I should able to. Give time. Don't read each and every answer of the students and then give the marks accordingly. Then only I will be far enough to provide the marks because simply. Providing me marks for pages are for the number of lines are only for key points that in my point of view won't work out properly. So I have to read the entire questions, answers written by the students and then. Based on that I'm supposed to give the marks. With this I can have the. That integrity in my question paper setting. As well as the assess.

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

Uh, yes, uh, sure if I've given the. Opportunity to work in VIP. May I know that possibilities of doing my postdoc degree? And the team with whom I am working.