Yeah. Hi, Arjun. This is Mavonika.

Thank you, Professor. Please start by introducing your professional journey in academia.

OK, uh, hi Arjun, this is Monica. Uh, I have done my. Bitex at UH. Jayme Kakinada and have done my Mtech at. Zen 2 Hyderabad and have. He has changed VIT value itself. I have a. Prove out all the US have been more interested in research. And. Have done a lot of projects and have a lot of publications, uh, mainly on electrical as well as an uh. AI and MLA areas. So now also I'm also working on uh. Project. I'm working as a researcher university. In a project like named as point to point autonomous.

So this is about my research and I have. Uh, I'm more interested related to the research. Where there is a collaborative work like electrical with computer science engineering methods. So it is a really interesting and it will be really useful for the further generations with smart work.

Interesting background, Professor. Thinking about your teaching experience, which foundational concept in artificial intelligence or multimedia do you most enjoy introducing to students for the first time, and why?

You recently published on attention networks for optimizing home energy management. When teaching this topic to students, how do you break down the concept of attention mechanisms in a way that’s accessible for those new to machine learning?

And later I have given a data set and a letter preprocessing has been done. And the model have been. Find 23rd. And uh. At the final we have got the output about accuracy was really good. I don't remember the perfect figure, but accuracy I have. What a really good one compared to other deep learning algorithms like LSTM, Gru. I have done a comparison analysis. So in such way we can teach. So we have to come from the basic and later we can go to the advanced.

Right, and when you compared the performance of your transformer-based approach to LSTM and GRU models, what specific metrics or validation strategies did you use to ensure the comparison was fair and meaningful?

Uh, OK, the other comes. Evaluation part of the Shirley bursting. So here, for each and every algorithm we come to printing analysis, there was. Certain performance metrics move. Used to call which are like RM missing. Open square error which we used to call. And also the accuracy. And also different factors are there based upon the problem of approach. So these are, uh, metrics we will be using for calculating the. Accuracy of each and every model.

Understood. When you introduce these evaluation metrics—like RMSE and accuracy—to students in a lab course, how do you help them interpret the results and connect these numbers to practical improvements in a real-world system?

Now you're asking about how I can, uh. In a practical part, that's a. It is really simple, so there everything depends upon the mathematical analysis. For that is the most important thing. So the math is very, very important. In this So for RMSC we'll have a formula will be there and for accuracy also. So the truth predicted values. Minus the true values, minus predicted values. We have to compare. Uh, and we have to evaluate the accuracy. So yeah, in that there is a mathematical expressions will be there, I will be. That will reach the students in the match itself. And we will give a. Google give certain sort of uh. Examples and will be able to talk. And later that, uh, we will be implementing in the practical. In the lab giving a program in the program for the same formula will be given. And it was easy for them because already in theory class we have. So just writing the code. That will be enough.

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

Support uh.

For example, for a prediction analysis. For RMSE will have the formula like. True values, uh, minus predicted. Like that being able to calculate the form. I think there is some sort of differences there in the formula. I am almost right I think. And later. For conditional analysis, this will be calculating and for another sort of. Machine learning models will be doing like. Confusion matrix also comes apart. So all the true values like are false positive, false negative, true positive and true negative. The diagnosis has to be. In such a way will be considering it and later will be calculating the call. Precision. Uh, there's some, some, uh, certain, uh, formulas are there. For that's what for each and every problem we'll be using some all these performance matrix.

Let's discuss your research trajectory for the next three years. How would you structure your research agenda to contribute to our university’s mission of academic excellence, particularly focusing on impactful publications and collaborations within your specialization?

Let's shift to your instructional methods. Suppose you're assigned a large-enrollment undergraduate course in artificial intelligence and cannot use traditional lectures or slides. How would you design an active learning experience—such as a flipped classroom or hands-on model—to engage and support all students effectively?

Jim, can you repeat it?

Of course. How would you create an engaging, student-centered experience for a large AI class if you couldn’t use lectures or slides—what specific strategies or activities would you implement to ensure active participation and learning?

To be honest, and it would be really like a boring part. So. Every if if the teaching normal class. Also if you do it in the lab. If able to parallel make them to do. That will be really interesting. So. Mostly I prefer hands on only. If, because it will be more quite interesting at least. Then students might not give 100%. At least. 10%. They might too. So that's what uh. I prefer.

His lectures are some. I I.

Understood. What’s one concrete in-class activity or project you would use in this AI course to help students apply concepts like neural networks or attention mechanisms without relying on traditional lectures?

You emphasized hands-on learning in the lab to keep students engaged—can you give a specific example of a lab activity or project you’ve used (or would use) to teach a core AI concept like neural networks, making sure students with different backgrounds can participate meaningfully?

Sure. During my PhD I used to go for Python. I used to go and use a Python classes. Like I used to do today. So at that time, uh. Along with the. Hospitals used to do. Very. They used to listen. Very interesting. Whatever the professor used to thought. And I also have learned and. Uh, of course I professor has to teach at like uh. How to? I mean. For example, what I can say.

How to use a logical operations at thematic operations with Python? And they are many things which how to calculate the length. Whatever the input, they are giving each and everything. The the professor has to tease them along with the. We I also used to learn. And the students by engaging in such a way. Students educate more ideas and. They in the robust manner they used to teach us. Also, Sir, why can't you? Uh, implement this sort of idea? Why can't this and why can't why the code is written by uh in such a way? What is the meaning of that address? And uh. That's sort of engaging views I have experienced during my PSD.

Let’s talk about institutional contributions. Suppose your department wants to improve the “research publications per faculty” metric for the next accreditation cycle. What systematic approach would you propose to help the department meet this target?

You mentioned engaging students by letting them suggest new ideas and approach problems collaboratively—how would you ensure fairness and accuracy in evaluating their performance, especially when students contribute very differently in group-based activities?

Of course, mainly this sort of thing comes under when we will be giving a projects to do. Minor projects and major projects will be given to the students. Of course, they will be forming a group and they used to. Have to do any group. So some students are very genuine and they will do it and some students they are there with a certain lack. Because of some distraction. So. The thing is, uh, if for such sort of students, uh. Its better to take a vivas out of the. Support if they give. At least because if we do, if we make a Viber present, we ask questions, vibas are sort of things, then it will be easy for them to. Uh, give the presentation. This thing will be helpful for the presentation part in the in that time of review. So definitely the professor if I'm a professor and if. I have a batch of such students, then definitely. Uh, every week or every, uh, whatever the. Based on the time span of the project, and definitely I'll be asking a Viber sort of thing. Because at least the basic things have to know. If not the project, it is the basic things they have to know. Which might be useful for us.

Thank you for explaining your approach to using vivas for group projects. When you conduct these vivas, what specific criteria or evidence do you look for to determine if a student truly understands their contribution versus just memorizing answers from teammates?

FIFA suppose if I had given. If they're using a Raspberry right, of course a coding part has to be deployed. In the Raspberry Pi for getting some sort of output. So of course it is a batch. One person might have done that. Another two members, don't know how. This person. So I can ask them. How in that code how this is? Our I will create my own error. I will. I will say that this is a syntax error and how you got. Specifically, the other schools were held. So this is the syntax error. How you got it and how you can rectify that? I intentionally do that and ask them make today. I know it was but. If it will make them to search. And. Whatever the AI tools, at least they know. They will be in search of finding a solution. So that's, uh. I'll make them do.

Let’s discuss a scenario around academic integrity. Imagine you are accused of grading bias by a student, but your department also expects a certain pass percentage to meet institutional goals. How would you handle this situation to ensure both fairness and alignment with the university’s standards?

Umm, that's really a good question. So yes.

You said that's a good question—can you walk me through the first step you would take to investigate and address the student's allegation, while maintaining both transparency and institutional standards?

They have to learn if not. Some for some students, uh, I'll give a chance for them. Of course I've I've lost HBO or whoever the hell is that, I will ask him to give him another opportunity to write an exam. Because some students who. By giving one opportunity, they they will really go. So at least we they deserve a chance. So I will ask them to to give a chance. To again write exam. And, uh, some might. Get Excel in that and some may not. So in that if. The person is excellent, then it's really good. I will ask them to consider this map. Then I can't do anything the the student has to face. Uh, because.

Even though if one chance has been given in. That is not utilized at management or. Even though he's OK with that, he also has a given chance. That has to be a pleasure. If they are not. Knowing that then. It's it's waste of time.

Let's discuss industry alignment. What concrete steps would you take to leverage your professional network and help create internship or employment pipelines for students in multimedia or artificial intelligence domains?

Now, I'd like to verify a few key qualifications for this role. Can you walk me through your structured approach to teaching a complex topic—how do you ensure clarity, logical progression, and engagement, especially for students who may be struggling?

If we give. Then it will be very useful for.

If we give. Then it will be very useful for.

You mentioned making time for students who are struggling—can you give a concrete example of how you adapted your teaching style or materials to help a particular student grasp a difficult concept?

You described giving extra time and focused support to struggling students—can you share a specific example where this approach led to a measurable improvement in a student's understanding or performance?

See in a lab, if they are 16, students are learning their Python cod. And everyone are getting the output whatever the professor is has given. Some students might not be getting and the. Of course, it comes to the majority of the students who have done the. And then 64 if 50 students have done the output and 10 students have been. At all, then we have to concentrate so. We have to hold them, we have to ask them to proceed further. Because 60 students in the class have to. Go in a parallel way. They shouldn't get interrupted because. Uh, if you concentrate on this ten students at that particular time, then 40 students might be living here. Show Vyasathan students. Uh, stop at that particular question at that particular time and have to. Make further. And after the time ends, like if suppose a lab is going on for two hours. So for two hours we have to make them go parallel.

After two hours. Those who didn't get out, what we have to make them set. For that particular question have to may ask them.

Right, so when supporting that group of students lagging behind in the lab, what specific techniques or adjustments did you make to the lab structure or your instructions to help them catch up without holding back the rest of the class?

Got it. For those students you set aside after lab hours, what method did you use to identify the specific misunderstanding they had—did you use targeted questions, demonstrations, or another approach?

The thing is, uh, first I will observe what they have done first. And where the? The thing is in the programming part there if you if you. There is any uh. Wrong in that particular line. The person who is sitting there. Have to understand the error first if he's not able to understand that error. Then definitely the person will stop the programming there. So we have to win the. Person to understand what why we are getting first. Understanding the error. Later we can. So I will ask them to understand the error, why they are getting the error but on what line they are getting the error and why will get the error. So in further if we get. The same error so that he can rectify by themselves. By not. Waiting for some other. So in that way I can able to help them.

Makes sense. Earlier you mentioned using AI tools and Raspberry Pi—can you expand on your direct experience with industry projects or consultancy, and describe a specific project where you applied your expertise to solve a real-world problem?

Uh, of course. Actually now whatever the project now I'm doing, uh, comes under this. Testing. Now I'm working on a project on Point to Point autonomous vehicle where. The vehicle needs to be designed. For mining purpose. So with your inner mining sector, of course there will be huge dust is there and there will be no signals will be there. So with the help of Lidars and some other camera sensors and, uh, microcontrollers like. Uh, these things I will be using, uh, we have to use. And we have to make synchronize. Because history has to be communicating with each other. And it has to detect the environment. And the lifter has to be moved. So this is ultimate aim we have to do first. So. So we before going to this major part, we want to implement for a small report. Abhiyaan, we want to design a small robot which is having a small 2D LIDAR. And the Raspberry Pi which we have have in the main controller. And we are designed a small report we will able to. Detect the environment with respect to the ladder so that it will, uh, move automatically. Whether whatever the path we have inside.

So that's what we are trying to.

Interesting application—when integrating the LIDAR and camera data for your mining robot, what method did you use to synchronize and process the sensor inputs in real time, and how did you handle noisy or inconsistent readings in that environment?

Interesting application—when integrating the LIDAR and camera data for your mining robot, what method did you use to synchronize and process the sensor inputs in real time, and how did you handle noisy or missing data streams?

Today we are in the starting now the project has been started a month back. So we are in a starting process itself. We are not. On indaponday. Now we are struggling in synchronizing the. Three things. So we are facing any errors. So. It will take time to describe in depth about it.

Understood. Moving to another key qualification—can you tell me about your PhD specialization and how your research contributes to advancements in multimedia or AI in media?

Background first. What have used AI and ML methods for implementing my house? Check doing my PhD. So that title of my PhD is A design of Advanced Optimization techniques in Home Energy management system. With by using attention models as well as the reinforcement learning tech.

So in home energy management system area I have used all these things. Of course, if that comes under electrical but. I have to. Have done reinforcement learning for optimizing the cost first. So the cost comes effectively here. The thing is, uh. Whatever the appliances are there. They have to be scheduled. With respect to low cost pricing?

You mentioned using attention models and reinforcement learning in your PhD work—can you walk me through how you adapted these AI techniques specifically for optimizing home energy management, and what results you observed?

Uh, they're able to schedule, maybe schedule loads like washing machine. And dishwashers. There's some there are some sort of schedulable load. Which we have rescheduled. At the low cost.

You mentioned using reinforcement learning to optimize appliance scheduling—can you describe the reward structure you designed, and how you balanced cost savings with user comfort or appliance constraints?

You mentioned rescheduling appliances like washing machines to minimize cost—what data sources or predictive methods did you use to anticipate low-cost periods for optimal scheduling?

And it has to be. Consider the user user. Because it doesn't mean that, uh, we can't neglect the user prefer. So that by considering all these factors. I have implemented reinforcement learning and I have done the first optimization analysis during my PhD research.

Right. For your PhD research, what challenges did you face when applying attention models to the scheduling of home appliances, and how did you validate the effectiveness of your optimization approach?

With comes to the data set. Because here, if you have in India, we don't have some sort of. Radiation. To have taken foreign data. And the foreign data set we have a different sort of pricing is there, but. The results out of lack of data is often tested. The missing values are there because every minute Samsung is connecting and the sensors won't work sometimes. So in certain days of the facing challenges and when comes to the programming part. These are pre processing techniques maybe. After getting the data, the RR program well. I'm. Writing the compatibility issues I have faced. Uh, all the things are. Happy is the challenges and when come to the validation. Of course I have to show that my algorithm what I have been implemented. So I have implemented the other algorithms, uh, manually, the advanced uh. By seeing all the research papers recent research papers I have gone through have considered them, I have done the. The algorithm part and I have compared it with. In such a way I have done the valuation and I have published all the. I have done many publications on my research. During my PhD. They are writing. Conferences as well as adjournments. SCI publications.

Thank you for explaining your validation process. Shifting focus, can you describe how you structure a lecture or laboratory session to ensure students with varying backgrounds remain engaged and achieve the intended learning outcomes?

Project integral is sort of area is there I will assign to him. And whereas a mechanical. And whereas the computer science engineering will come in a programming AI, machine learning. They can implement so all together. Finally all together comes under 1 system. So that. Engaging it will be really cool. So that Asian every student will know the other person. And. While combining their face also some some sort of issues and they will also. Try them to engage with others. How to rectify this? In such a way you can.

You described assigning project components based on student backgrounds—can you share a specific example where this approach helped resolve a team conflict or boosted overall project quality?

Follow the decorations.

You described breaking down projects and assigning roles to students from different disciplines—can you give a concrete example of how you ensured each student understood the connections between their component and the overall system?

Yeah, yeah, yeah.

Understood. For your experience in industry projects or consultancy, can you walk me through a specific instance where you applied your research or technical skills to solve a real-world problem outside of academia?

Search academia now I'm working. Presently the previously I was uh. My complete academic order continues. Letech, Amtech PhD. Now I have started doing the project. So, uh. Previously I don't have some sort of experience now. I'm working on.

Okay. Since your industry-facing work is just beginning, can you explain how you plan to apply your reinforcement learning expertise to the current project you mentioned, and what a successful outcome would look like for you?

Pause RL will be implemented but.

There is a many sort of complexity welcome.

Understood. Since you mentioned you're beginning to implement reinforcement learning in your current project, what specific challenges do you anticipate facing in terms of data collection or real-world deployment, based on your previous work?

While implementing in the RM because. The vehicle we are designing is. The mainly the purpose of using it is in the mining. So in that mining areas.

You mentioned anticipating complexity in implementing reinforcement learning for your new project—what is one concrete strategy you plan to use to manage or simplify the integration process as you begin?

RL uh, whether I have to check whether it can be implemented or not? 1st if it is implemented then that will be really gonna really agree. Uh, because RL. Welcome Chandler to the controlling part only. But as there are many, uh. Census are there so while integrating how this RL will work or not? I have to check it first.

You mentioned designing a vehicle for mining areas—what is one specific constraint from the mining environment that influences your approach to reinforcement learning in this project?

Since you mentioned sensor integration as a challenge for reinforcement learning in your mining vehicle project, what approach will you use to test and validate that the RL system is actually making effective control decisions given unreliable or noisy sensor data?

Show here in the vehicle the controlling part comes under steering. Cheering as well as the break here are the two controlling where the RL we can able to. With the help of RLCB.

You mentioned using RL for steering and braking via RLCB; can you describe how you would determine if the RL agent's decisions are both safe and reliable before deploying on an actual mining vehicle?

So that we have to check.

So that we have to check.

Okay. You mentioned steering and braking as control points for RL—what metrics or signals will you use to judge if the RL agent is actually improving vehicle performance during your initial tests?

Understood. Shifting focus, can you explain how you ensure your lectures and laboratory sessions have a clear structure that helps students follow complex material, especially when introducing topics like reinforcement learning?

Sorry Arjuna, I didn't hear you. Can you explain it?

Sorry Arjuna, I didn't hear you. Can you explain it?

No problem. When you introduce a new topic like reinforcement learning to students, what is the very first concept or example you use to capture their attention and set the stage for deeper understanding?

Sorry Arjuna, I didn't hear you. Can you explain it?

Sorry Arjuna, I didn't hear you. Can you explain it?

No problem. When you introduce a new topic like reinforcement learning to students, what is the very first concept or example you use to capture their attention and set the stage for deeper understanding?

No problem. When you introduce a complex topic like reinforcement learning in a laboratory setting, how do you design the first hands-on experiment to ensure students grasp both the theoretical and practical aspects?

No problem. You mentioned integrating RL with sensor data on mining vehicles—when you teach similar advanced topics, how do you break down the complexity so that students with different academic backgrounds can all follow along?

What they call you can able to apply the brake. So this comes under everything under the mechanical, uh, student part. And when it comes to the. Electrical thing, uh. So each and every sensor, the power supply and instrument. The power electronics comes under picture here. So that supply needs to be given. So buck boost converters are are gonna be used here so far that. May have to give a supply. So fathat electrical, uh. Background Students. Within Peaceful, yeah. So this three students have to. Work together. Show Dhanuri Dakota will work.

You mentioned the importance of collaboration between students from mechanical, electrical, and electronics backgrounds on your project—how do you facilitate clear communication and structured teamwork among such a diverse student group to ensure the project moves forward efficiently?

For suppose in electrical part. The system has to be designed for the bubbles. Converters need to be ready first, appliance needs to be given. And the LinkedIn it has to be. List all the converters, the cameras. Has to be placed in mechanical part. So the first quarter, the things we required, we have to be analyzed first. Then based on the the vehicle needs to be designed. You'll come some again. Show the mechanical student, whoever is there, whatever the electrical part that has been designed, he will be able to solve the thing. And the vehicle will be designed there. That, uh.

In what example I can explain to?

I'd like you to work through that yourself, but let's move forward. Based on your academic path—BTech, MTech, and PhD in succession—how have you ensured your teaching approach remains structured and clear, especially when guiding new students through foundational concepts?

In that way, the students can go. And later. The.

Understood. When you oversee student teams designing both the mechanical and electrical systems, how do you evaluate each student's individual contribution to ensure fair assessment during project grading?

Okay. When it comes to guiding student research projects, can you share a specific example of how you've helped a student refine their research question or methodology to improve the quality of their project?