Good day, Shweta Anand Redkar, Ph.D. I'm Arjun. I've had a look through your background and your professional record. I'm interested in hearing more about your work and how you see yourself fitting in here with our faculty. Ready to start?

Please start by introducing your professional journey in academia.

OK, good morning. Uh, my name is Doctor Shweta Redkar and umm, I'm uh, having AB in uh, information technology and uh, masters in computer science and Engineering.

Uh, I've done my PhD in UH computer data.

Data science and computer applications, my pho topic is uh, intelligent prediction model for drug repurposing and umm, it's been uh, eight years I've been in teaching, uh, industry. I was working in Goa, uh, in an engineering college for almost for eight years, five years and in Manipal University, Jaipur for almost for three years. That becomes total eight years and. My machine learning. My expertise is machine learning.

Uh, then umm.

Object Oriented programming, then RDPMS. So the subjects which I've taught. Besides that I have three journal papers, 2 are SCI and one is ESI. Then I have around 5 Scopus indexed conference papers. I have 3 patents which are already published. Besides that I have.

Interest in mentoring students which are usually do and I usually do outreach activities like offering save for longer in Gurudwara and then for other organizations like Isha Foundation where I volunteer for them whenever at state level whenever such programs happen at at that place is.

So this is what my introduction is all about.

Thank you for sharing that, Dr. Redkar. I appreciate the detailed introduction regarding your academic journey and contributions.

Professor, let us begin by assessing your expertise in Artificial Intelligence and Machine Learning. Could you discuss how AI and Machine Learning methodologies are applied in healthcare, and provide an example where this application has led to substantial improvements or breakthroughs?

My PhD work is, uh, regarding this, uh, area. So my PhD work is, I have already said that, umm, I have a PhD in, uh, drug repurposing where I'm using machine learning techniques to identify a drug which has already been failed, which is already in market and which has been failed. I'm utilizing this drug to, uh, find out a new indication, uh, and uh.

The models that I have worked on have given good results and the predictions what I've done are around about 90% and above.

Results I have got where it can show that over these other drugs which can interact with proteins and for this I have used various machine learning techniques like SVM then.

KNN then uh, Random Forest XG Boost. So these are the uh models which I've worked with and umm, it is seen that it is.

It has given good performance.

Understood, Professor. Thank you for elaborating on the methodologies you've employed. A follow-up question: given the performance you've achieved, how do you ensure the robustness and reliability of the results, particularly in the context of sensitive healthcare applications? For instance, how do you account for potential biases in data or ensure interpretability of your models?

OK, so the data set that I had used, uh, so I taken data from various websites where the pharmaceutical company will deposit their.

Findings.

They will, uh, deposit their findings and, uh, I generated my own data set, uh, for testing purpose. And, uh, after doing the from, uh, and the for testing purpose, I had used the data set called binding DB and from there I had tested my new entries with the existing models which are already trained.

And uh, I've seen that, OK, they have given a better results.

Thank you, Professor. That provides clarity into your approach to validating robustness. Last follow-up: in your experience using these models, such as SVM or Random Forest, how do you navigate the trade-offs between computational complexity and accuracy, particularly within time-sensitive healthcare settings?

Could you please come again?

Of course, Professor. Considering the machine learning models you've used, like SVM or Random Forest, how do you balance the trade-offs between computational complexity and accuracy when working in time-sensitive healthcare scenarios?

OK, so about this I have.

Was to remove the overfitting issues. I have used the key fold cross validation technique and make sure that none of the entries of drug and protein interactions lie in any of the fold so that I get a good so that there is no biasness while training.

So when I produce a new test data set, so it had given me accuracy about 93 and 95% at an alarm.

Understood, Professor. That thoughtful use of k-fold cross-validation undoubtedly strengthens model reliability. Let us now examine your ability to teach theory and laboratory courses. How would you structure an introductory graduate-level course combining foundational theory in AI and practical laboratory sessions for applications in healthcare? Can you outline key topics or modules?

OK.

So key topics could be over here, I am introducing the students with us. If, if they're working with proteomics, then introducing them with, uh, proteomics, the basics of proteomics and uh, navigating them through the uh, uh, basic level of fundamental levels of amino acids, then how protein chains are formed, then how sequences are read and interpreted that how UH-3 dimensional structures of the protein is interpreted and what are the various tools existing.

Access this existing material which is already deposited by the pharmaceutical companies and besides that, how to apply machine learning techniques, how to generate the data, how to preprocess the data or how to form your data sets that could be told and then later on apply the machine learning or deep learning techniques.

Or you can use to.

Thank you, Professor. Your structure provides a solid foundation that integrates domain knowledge with hands-on application. As a follow-up: how would you evaluate the effectiveness of students in grasping both the theoretical concepts and practical applications during such a course? Would you prioritize assessments, projects, or another approach?

To understand whether they have fully understood the practical aspects of it. So, uh, I would, uh, continuously, uh, assess them using some components such as quizzes or assignments or. And besides that, I will also give them some projects to understand whether they have fully or thoroughly understood the concept or not.

OK. I would prioritize both of them.

That approach is balanced and ensures comprehensive evaluation, Professor. One final question on this: in laboratory sessions, how would you guide students who struggle to bridge the gap between theoretical concepts and hands-on applications, especially when working with real-world healthcare datasets?

OK, so this could be done. It's like your I need my students help. So if they are not practicing by themselves and waiting for the last moment to study, then it is difficult for them to understand the and cross the concepts. So as I've said that OK, there will be continuous evaluation in the every month they might be evaluated whether they have understood or not why acquisitions or assignments. So we need to see whether they have fully understood.

Not if they have not understood, then I would, uh, mentor those students, those who are, uh, uh, having low performance, like slow learners. And, uh, I will, uh, bifurcate this slow learners and advanced learners. And if advanced learners are there, I would, uh, surely like to go ahead with them with some research work and give them some research topic to work on so that we can go ahead for some publications or we can write a solid project. And those people, those who are those students, those who are low performers, OK, having low performance.

Uh, they will will be giving some, given some more test to work on or some more, uh, uh, hands on to work on. That's how you can bridge the crap there. There is no other way to bridge the crap.

Thank you for this clear strategy, Professor. It demonstrates a strong commitment to tailored mentorship. Now, let us move to your experience in student evaluation and exam duties. Could you detail a method or approach you would use to ensure fairness and reliability in assessing student performance during exams or other evaluations, particularly in technical subjects like AI and health informatics?

Huh. There is some, there are various techniques where we can use to evaluate the students. Like I will go ahead with the Blooms technology taxonomy where we are accessing whether the student has understood the basics, then whether the student has understood the concept and whether they can. Evaluate themselves and apply that topic somewhere else in EI and artificial using AI and in the healthcare.

So the technique would be the similar technique over there as well, uh, continuously assessing the students and via test or via group discussions, we can identify whether the student has understood or how much that person has understood or not.

Thank you, Professor. You’ve presented a structured approach that integrates Bloom's Taxonomy effectively. As a follow-up: how would you handle scenarios where a portion of the exam results or evaluations shows unexpected bias or inconsistency in grading? How would you address such discrepancies?

I do not get this question like what exactly are yours? Like what exactly this question means?

MMM.

or inconsistent criteria, how would you resolve this issue to ensure fairness across the board?

OK, so since we are already having a question paper which is already been set and certain questions are having certain marks.

So the questions might have certain questions will have steps and if they have followed those steps to.

Measure how far the student has understood the concept and you can award them marks. So I don't think so if you are setting your paper in this way or question paper in this way, there is no like, you know, biased corrections.

Reach to an outcome, or to reach to a result, or to reach to an answer. Those steps will be having markings so from that you can.

Understood, Professor. A stepwise marking scheme indeed helps maintain objectivity and fairness. Transitioning now to your ability to guide student projects and research, could you illustrate how you would mentor a student working on a machine learning project in healthcare, ensuring their research stays innovative yet practical?

OK, currently we are working. Currently I'm working on some projects with the students and with my own students where where I'm currently working and I've given some data sets to them. So already they are being told what the data set is about and what is this topic about and what the outcome would be. So they have fully understood the problem statement. So now since the data set is already made them available.

And they're fully understood, uh, how the data set, uh, the description of the data set. So they are, uh, trying to work on the, uh.

Preprocessing part and once the preprocessing part is done then they would be applying their machine learning or deep learning techniques or soft computing techniques.

Thank you, Professor. Your methodology of guiding students through datasets and problem statements ensures a practical learning path. As a follow-up: when students encounter challenges, such as missing or imbalanced data, how would you mentor them in addressing these issues effectively?

OK, missing, uh, related issues, missing values that could be addressed in various ways like, uh, computing the key means, sorry, uh, not K means, uh, computing the mean value or mode value or median value. There is another way to compute this missing value like a K&N imputation. So that's the way we are going to preprocess the data. And about imbalanced data, yes, the healthcare data is imbalanced in nature.

Or random under sampling to balance out your data set. Uh so that we can use it for training purpose.

So we could use various algorithms such as smoot.

Thank you, Professor. Your solutions for missing and imbalanced data demonstrate excellent technical insight. Let us now move to assessing your communication skills and approach to structured teaching. Could you share how you present complex topics, such as a deep learning model's architecture, to a diverse audience of graduate students, ensuring clarity and engagement for all?

OK, so about the topic, before introducing them the topic, I would just brief out them on the topic like what this topic is all about, why I why are we studying this topic and what could be the various applications for various algorithms. So once they are aware about why are we studying this particular course and what is the.

Reason to study this course. Then I introduced the topic. I stressed from the basics so that you know, I give them very simple examples 1st and then I move on to the complex examples like I straight away do not tell them OK, this is the handwritten data set available or this is the Titanic data set available. So from where you have got this data, how you have got this data? And what is the way to collect this data then? It's it's not that only data is available and then you.

Start working on it. You need to preprocess that. So what are the various preprocessing techniques available that I would be telling them Then if any feature selection is required, if you are using machine learning technique, then I would be making them aware about that and if it is deep learning then there is no feature selection involved. So it's also depends on what application that you're working on, whether it is image or whether it is sequence. Based on that we are going to select the algorithm for deep learning purpose and how the training is done for these models and. How to get the results? How to interpret those results? What does that mean?

And how you can use this algorithm besides the data set that is which has been made available to them to other data sets that could be assessed?

Thank you, Professor. Your strategy of building foundational understanding before diving into complex examples fosters clarity and engagement. A follow-up: how do you ensure inclusivity and address diverse learning needs among students with varied skill levels or academic backgrounds in such sessions?

Could you please repeat that question again?

Certainly, Professor. How do you ensure inclusivity and accommodate diverse learning needs among students with varying skill levels or academic backgrounds when explaining complex topics like deep learning architectures?

How to make that complex topic simpler to them? How? How do I tell them how to make it simpler? Is it is? Is this the question means that?

Are you asking me about uh?

OK, so in the beginning itself or if I have a class of 60 students, the strength is 60. So in the beginning itself it is not possible to identify which student has understood or not.

Yes, Professor, that's correct. I'm asking how you tailor your explanations to ensure that students of different skill levels or backgrounds can understand complex topics effectively.

So I would be discussing with them as I've said that I would be.

Introducing the topic and telling them in a very simple language. Then why are we studying this topic and for what? What could the applications? So I'm going to ask them what are the applications? What is coming to their minds? First I would give them some application and then I would ask applications from them whether they have understood the topic or not.

That we are going to cover next.

Now says opposed even after giving the entire explanation, and they still do not understand this. As I've already said earlier, there might be slow learners and there will be advanced learners. Advanced learners will be picked out, uh, based on certain, uh, criteria. Like, uh, I might put a threshold to assess them and who are above that threshold, I'll select those students for advanced topics like research, publications and all, and those who are below that. Below students.

So below students, it doesn't mean that students are bad. Their clustering level is different, everybody's. So below students will be given some different example, OK, if they're not understood the same example, they will be given some different example to understand and that's how things will be made simpler.

Understood, Professor. Your approach of adapting teaching methods and examples to students' levels ensures inclusivity and engagement effectively. Let us now move to verifying your qualifications—your PhD specialization aligns with AI and healthcare applications. Could you elaborate on how your doctoral research has prepared you for leadership in guiding institutional research initiatives or collaborations in these domains?

Sorry to interrupt, we're down to the last 5 minutes.

OK so my doctoral research is related to healthcare and I had chosen this topic which is related to giving something back to the society. So here I'm taking some drugs. Where, as I've already said, the drugs are already failed. So we cannot throw these drugs out from the market because the companies have already, you know, spent so much money in billions.

Of while manufacturing this, so I'll use this, uh, drugs, uh, for, uh, identifying new certain indications if they are existing and we can, uh, conduct the vet lab studies for this.

So definitely our projects can be written and currently I'm writing some projects where we can do some funding work, where I can get some funding. And I can, as I've said, I already have the advanced learners and I'm expecting that the organization that I'm appearing for interview will be providing PhD students so I can give this PhD topics to them as their PhD thesis work.

That's how we can work on this topic.

And.

Thank you, Professor. Your doctoral research and proactive approach to securing funding demonstrate significant potential for leading institutional research efforts. Moving on to your research publications, could you provide an overview of your most impactful publication in AI or healthcare, highlighting its significance?

Yes, I have one publication which is already there in SCI journal and which is talking about how healthcare can be, how AI can be used or machine learning can be used in healthcare systems. The 2nd paper that I have in Spiritual Nature in Scientific, sorry Nature in Scientific reports and that is talking about how effectively we can use heterogeneous servers.

To allocate the activities so we can use this in healthcare system like for example if there are patients in a hospital where we can allocate the resource like for example doctors and nurses and what voice to the most.

Person who significantly or desperately want the resource so this could be done.

You know the most.

So these are the papers that I have.

Thank you, Professor. Your publications reflect both technical innovation and real-world applicability in healthcare. As a follow-up: how do you envision leveraging such impactful research to foster collaborations with industry or academic peers for advancing healthcare technologies?

We need to identify certain pharmaceutical companies where we can collaborate with and that identification need to be done. And as I've already said, OK, funding can be applied. So we need to identify which.

With industry.

Company Weekend collaborate with.

Because the drugs are having their own license, so we need to identify that.

Understood, Professor. Identifying suitable industry collaborators while securing funding is indeed crucial for impactful research. Lastly, regarding your experience with industry projects or consultancy, could you share an example where your expertise in AI and healthcare directly contributed to solving a significant industry challenge?

Currently I have no projects in industry. I'm writing the projects.

Thank you for clarifying, Professor. That concludes the verification phase of your skills. Let me confirm the next steps internally. One moment, please.

Professor, let's begin with your approach to teaching. In an introductory course within your field, what strategies or methods do you find most effective for engaging students in foundational concepts?

Could you please repeat that question again?

Certainly. In an introductory course within your field, what strategies or methods do you find most effective for engaging students in foundational concepts?

With respect to machine learning or AI?

With with this.

With respect to machine learning or EI.

Or healthcare.

Are you asking me about that?

So if it is AI or healthcare related or predictions then we definitely need to know Python for this because for machine learning we require the language which is supported for machine learning is Python.

That's a good point about Python being essential for machine learning applications. When you teach foundational AI concepts or predictive methods in healthcare, how do you ensure students grasp both the theoretical underpinnings and practical implementation using Python?

Will be done using.

Practical approach in theory or the theoretical?

And besides theory, it will be a practical approach where you give you start with basic examples like a + b how variables are stored.

Explanations will be provided to them regarding Python.

And how you get to see the output, what installations needs to be done, how to install that software and how to use that software to write your codes. So we'll start with basics first and not straight away with machine learning. So if the student is able to understand how to work with the interface of Python, then slowly, slowly how to. Incorporate files? How to import files that would be told? What are the barriers Libraries. Libraries used?

Uh, to, for importing purpose, what are the various data sets, how they can be, uh, imported and what are the various libraries for cross validation or for visualization purpose that would be told to them. Once that is done, then, uh, how to train a model, how to write the code for that for training a model that would be told. And then how to pass your test data set, how to, uh, test your data against the train model, how to see if your model that everything will be true to them.

There's only one way to assess this, whether they are getting output or not. So if they have got the output, they have done the work perfectly fine.

Understood. It sounds like you emphasize a structured progression, starting with Python basics and gradually advancing to machine learning workflows, including importing datasets, using libraries for visualization, and training/testing models. How do you assess whether students have successfully grasped these concepts and are ready to transition to more complex AI applications?

Yes, that is been already answered, uh a few minutes back.

Thank you, Professor. Ensuring students achieve the expected output is a clear approach for evaluation. Beyond producing the correct results, do you further evaluate their understanding of the conceptual process behind the code and its implications in real-world applications?

Motivation was, as I've already said, I wanted to give some back, something back to the society. Your PhD work has to give something back to the society and it should not be restricted to just writing the thesis, printing the thesis and submitting it to the university and getting the degree. So here I'm trying to predict something which is already failed and to manufacture this drugs, pharmaceutical.

Professor Redkar, I see that your research includes a project titled "An intelligent prediction model for drug repurposing using soft computing techniques" as part of your Ph.D. Could you briefly describe the motivation behind this work and the core achievements you've developed?

Spend around around 12 to 13 years for a drug to come up.

Out in the market and they also spend billions of dollars for this drug. So it's a huge investment in terms of time and in terms of money.

So.

We are trying to target to automize this process and this is what the significance and motivation was.

Understood, Professor Redkar. Automating the process to repurpose drugs certainly addresses the substantial time and financial challenges faced by the pharmaceutical industry. Considering this research, could you explain one key methodological approach you used in developing this prediction model, specifically in leveraging soft computing techniques?

OK, uh, soft computing techniques, uh, I have used, uh, certain feature selection, uh, methods, uh, like what could be the various features, uh, where the drugs would be binding to a particular protein. So for that I had used some top, some feature selection methods where I'd use genetical polythene swarm optimization, But, uh, what I have understood is, uh, and besides that, I'd use some statistical methods.