Good day, Dr Krishna Kumar L. 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

Good afternoon.

Uh, I introduce myself myself, Doctor Krishna Kumar.

Completed my post graduation and.

Doctorate in computer science and engineering in the area of specialization in artificial intelligence.

University.

PhD graduated in 2024.

And masters in.

2008.

Uh, this is my professional background and my work experience are in, uh, in a private college I have worked for from 2008 to 2021, uh, in the department of computer science and engineering in a different designations, uh, working for various NBA and NAC committees, teaching various.

Uh, subjects in the.

Department of Computer Science and Engineering.

Thank you.

You're welcome, Professor. Let's proceed. Could you elaborate on your experience with teaching subjects in Computer Science and Engineering? Specifically, which subjects have you focused on, and how do you approach ensuring students grasp both foundational and advanced concepts effectively?

Thank you for the question.

From my I have joined my.

Career in the year 2008.

Uh, from there onwards, from uh for 2008 to 12, nearly four to five years, I have a teacher data structure subject in UG level and PG level.

Uh, after that operating system, I have thought for two to three years.

With the theory and practicals.

Uh, from 2021, uh, from 2016 onwards, I have, uh, working on AI department, uh, there onwards I am teaching the foundations of artificial intelligence and ethics in artificial intelligence. And after that, the AI search techniques and, uh, few concept in machine learning and.

One main important subject, I thought in the artificial intelligence department.

For uh, advanced courses is uh, game theory in AI.

These are all the major courses which I have taught to students in UG level, PG level and Advanced level also.

Thank you for providing that overview. Could you elaborate on your approach to blending theory with practice in these subjects? For instance, how do you ensure that your students not only understand the theoretical foundations but can also apply these concepts in practical scenarios?

Thanks for the question. UH, even though I have started my career in 2008 as a fresher without having any uh teaching expertise, but uh in my post graduation and in my PG from my UG itself, I started teaching and briefing uh the concepts to other friends. Uh, in my understanding, uh, what I had is uh.

My knowledge improves by teaching. That is what I understand from my early childhoods. When I, uh, explain it more to others, I learn more so that I started teaching like that. So I am interested in teaching. That's how my passion also and my career journey also started. So to teach that students to make it understandable in theory, practical.

Uh, I will keep. I will give you a small instance which happened.

In my class so that it will be more familiar.

Uh, I was explaining on uh, data structures. Uh, I have worked in uh, different uh.

Uh, regions in a geographical regions in India, uh, so, uh, in different 2, two different various locations, I will expect, uh, in Tamil Nadu side, I were working, I am explaining A stacked data structure. OK, it is the pile for there. There I will be explaining the stack operations after that I will.

Put a CD box and from that CD box I will ask the students to take that CD box.

Access and or after that also I will ask them to insert and I will ask them to.

Pop so that they will understand the basic operations with the theory and practical why that analogy and I will ask a girl students to come over the stage and put a Bengals so that it will be easy for their understanding.

After that this is for the stack operation. So in that geographical location I told in the South Indian side when I was working in the North India or somewhere this middle of India region, there mostly the sandwich and bow bodies are more famous. So I have kept the pav bhaji as an example. I opened the pav bhaj and I keep the waters inside and layering of that and closing it and eating. So this is how.

UH demonstrated uh the stack operations with the theory and practical for the student understanding purpose. This is for the UG basic level.

Have I answered the question?

Yes, Professor, you've provided a clear and engaging response. Your use of relatable and region-specific analogies demonstrates an effective teaching strategy, making complex concepts easy for students to grasp. Thank you for sharing this approach. Let’s continue—can you elaborate on how you guide students in understanding advanced AI concepts like game theory, and how you encourage their involvement in research or independent projects in these areas?

Uh, game theory is, uh, thanks for the question. Game theory is a subject where it applied from, uh, ancient days to today. Uh, different, uh, government organizations, uh, different marketing areas, one to one. Uh, for example, AAA is having an interview with me and I am here, we are having a negotiation. This is also a part of game theory.

Uh, so I want to explain this game theory concepts to students. I have thought it to two cycles of, uh, undergraduate level, uh, students, uh, they're not under there for the AI students and they are familiar with the prerequisites that with the foundations of AI ethics in A and AI search techniques. All these are the cyclic courses and they came for the advanced course for the game theory.

In yay specific to this, my subject game theory deals with the approach of solving the problems using the gaming technology. I have thought many examples in that I am very much interested in the one specific concept called as prisoners dilemma where the police will have a strong investigation ideas. Between the.

Two prisoners who will make a confiscation and non confiscation. So I will ask a student to perform an activity based learning. So I will have one product that has been hided or is taken by some robbers and after that I have got two or three.

Students making them to make forcibly convict to that robbery.

As like as uh, police is doing, I will bring them to investigation area and I will make inquiries. So, uh, the seeing audience will understand that. So police is doing like this after a few time, uh, few minutes later on only they will understand that, uh, one person inside the room is giving 1 confiscation and the other person giving a different form of investigation. So while we are the, as an investigator, I am understanding both of the two. Prisoners, uh, ideas And after that I will make conclusions if both of them confess, what is the conclusion or addition I can make if one person confession and one person not confession. And how I can make a decision if, if it is a A&B, how much possibilities, A probability I can derive with reputation without repetition. These are all the things, uh, I will make it as an activity based learning so that the students will understand this prisoner's dilemma.

Or.

Very easily because they're prerequisite. Also it states that the foundations of AI ethics in AI and AI search methods they have studied and coming so that it will give some the good understanding of game theory when I make an activity based learning.

Moreover, this many other game theory topics also I have thought one of my interested topic I have explained.

What's my meeting list?

Do I get my transcript over here? I have not getting any transcript displayed here.

Yes, Sir, you can continue.

I cannot assist with checking your meeting list or transcripts. Please connect with the appropriate support team or platform administrator for assistance regarding these concerns. Shall we continue with the interview?

Good. You've shared detailed insights into your teaching methods and examples. Let’s turn to your research experience. Could you discuss the focus of your PhD research and any key findings or contributions you've made in the field of Artificial Intelligence and Data Science?

Thanks for the question. Uh, I have started my research, uh, in the year, uh, 2015 and 16. Uh, and hearing that, uh, yes, uh, when I have joined, uh, my, uh, after completing my post graduation in the career, I feel that I have to upskill myself and I have identified the gaps when I am teaching the PG students, the courses are getting advanced.

The areas of cloud computing distributed and slowly transforming towards AI and advanced computing technologies. So I am.

Uh, focused on artificial intelligence area, uh, which can be more helpful at that time there, it is not much, uh, light on AI, uh, there are problem statement. What I have identified is, uh, how I can make.

Vehicle communication more prominent and more easy. That was my PG results. Also adoc vehicle networks. That was my, uh, PG research.

So from that part itself I can take it forward. Now I am finding out the more problem statement or latest problem statement in the 2016.

From there I started my research. After that slowly it is transformed into vehicle communication and after that it has transformed into the autonomous vehicle communication where AI plays a vital role in decision making of autonomous vehicles. It plays a vital role that how the communication happens between a vehicle to vehicle.

In decision making, vehicle to infrastructure for decision making. So these are all the major areas and my contributions in various publications and it was listed in my.

Resume also.

My major contributions if I said in my PhD it is the vehicular autonomous vehicular communications using AI and machine learning technologies.

And published various journals.

Yeah.

Anything you want to add more?

Certainly, Professor. As you’ve highlighted your contributions to autonomous vehicular communications, could you also reflect on how your research aligns with current advancements in AI? For example, are there advancements like reinforcement learning, edge computing, or federated learning that you’ve utilized or foresee as pivotal for this domain?

Yeah. Thank you for the question. Uh, yeah. When I started my research, there was uh, uh, not much, uh.

Uh familiarity on uh autonomous vehicle communication in uh twenty 20s in late twenty 20s only the vehicle communication, uh, autonomous vehicle communication has been more initiated. Uh from 2016 onwards, I am working on the vehicular communication that is the ADOC networks. That was my major research.

After that once I hear visualizing that more research is happening on the autonomous vehicle research. So I find out that where is this autonomous vehicle is working on currently. So I have seen that most of that already existing autonomous vehicles are in the military or defense or in.

Uh, flights, uh, that was, uh, working. So, uh, I have started working on, uh, studying on, uh, and researching on what are all the research available for autopilot modes available in the aeroplane or flights, uh, from there I got an idea that, uh, Baylor circuits.

Uh.

Uh plays a vital role in making a decisions in autopilots. From there I have derived the concept uh, Ted boiler circuits, uh, uh, so that I have integrated that in the, uh, vehicle to vehicle communication that ways that that happens between air base to autopilot mode in the flight.

So how this system can be adapted to the?

Roadside vehicles such as the car to car or car to infrastructure. Most of my research concentrated on autonomous vehicle. What I deal with is the car.

So how it can be, uh, the problem can be easily solved? I derived the few concept, uh, that is the main core area where I, uh, working on dead boiler circuits. Uh, so I have taken that approach and I have insisted that in the research of vehicle to vehicle communication and vehicle to infrastructure communication.

There I find that the more, uh, latencies.

Happening between the communication. So there, uh, there was a, a drawbacks also, I could able to find and, uh, to remove that gaps. I were working in advanced, uh, advancement of AI. That is the boiler circuits were used for, uh, incorporating the communication and AI will, uh, have, uh, more. Uh, decision making approach in vehicle to vehicle and vehicle to infrastructure communication.

Understood, Professor. Your integration of Baylor circuits with AI for autonomous vehicular communication demonstrates insightful adaptation and innovation. Addressing latency in communication is indeed critical, and it’s commendable that you've tackled these challenges. What specific AI advancement or technique did you find most effective in reducing these latencies and improving decision-making accuracy for vehicle-to-vehicle and vehicle-to-infrastructure interactions?

We are mostly on concentrating on vehicle to vehicle communication.

The communication happens or else the decision making happens in the.

Uh, what uh, the, uh, data center where it is generated like age, the vehicle understands the images there itself or it grabs the videos or it gets the information through ladders there itself. It, it will pass that information or else the data to the infrastructure and infrastructure also will make some predictions and give it to.

My vehicle and the vehicle itself also has to make some decisions there deep learning approach that is the in neural network we will be saying many hidden layers will be there. There will be 3 hidden layers.

Destination layer in between 3 hidden layers has been created. That hidden layers will make the dish nest like as a neural network, uh, so that it will fine tune the decision and immediately the whenever it is needed to accelerate break. And in most of the autonomous vehicles, we will try to avoid the clutch system so that accelerator and brake only we have to be concentrating.

UH-1 is the source UH layer and last one is the.

Uh, to make more, uh, accurate and, uh, precisely the decisions, yeah. Uh, deep learning with neural networks has been integrating in solving this problem.

Thank you for that detailed explanation, Professor. Your integration of deep learning and neural networks to enhance vehicle decision-making by leveraging edge computing and layered predictions is impressive. You've mentioned the use of multiple hidden layers in the neural network for optimized decision-making. Could you share any challenges or limitations you encountered in training or deploying these neural network models in real-time vehicular systems, and how you addressed those?

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

Thank you.

Can I answer?

Yes, please. Go ahead, Professor.

OK. Thank you.

The problem what I faced in neural network or deep learning is when we have a large amount of data that is the letter communication will gives us a large amount of data. That large amount of data nearly 8 sensors, front vision cameras, rear cameras, side cameras, all these input data has to be.

Analyze it on the spot and we have to find out the decision for that. The image has to be interpreted and converted into data model and from later from sensors I will get information that is that will be in the text format data.

So video has to be interpreted and converted into uh, text, uh, images has to be converted into text and the sensor data is in the text format. All these three data has to be collected first. This is the one uh, one side of that collection node. And after that I told another hidden layer that is the first hidden layer, what it will do. It will understand the uh text input data, what it is used for or how the data sets are available and.

Tuning of the datas, uh, and it will be in the first layer. After that, in the second layer, it will go for the training. All this will happen in the, uh, milliseconds fashion. Then only the autonomous vehicle can make uh, decisions promptly and precisely. Excuse me.

Uh, that, uh, you know, firstly, after passing that information from first layer to second layer, from second layer, what we will get already existing data. We have taken a a large amount of uh, training data set that has been available in.

USA and UK government datasets. In Indian scenario there is no such large amount of letter datasets. So we have taken existing Lidar data set from other countries and we have.

Making a testing. So that existing leader data set will be.

Taken and this layer 2 input will be provided to layer one input will be provided to layer 2. So there will be a large amount of comparison will happen whether this data is enough for making a decisions or not. It will go to the third layer where the exact decision is taken with the prompt existing available.

Precisely the data, uh, which is, uh, in front side, there is a, for example, I'm sorry, uh, saying inference side, there is a, uh, trespass, a human uh, is coming and backside there was a, a wall or something is uh, or else in another vehicle is there the car cannot move forward also backward also inside it will understand what are all the space available so it can take take the capture that Lib said camera on right side camera.

Uh, vehicle, uh, steering angle has to be rotated and it has to take little right or left. So can it can, uh, pass through easily. Uh, these are all the main uh, uh, works of these 3 hidden layers and it will add the final stage. It will go for the another output layer. So these 3 hidden layers work like this.

And it will identify that it, uh, once that friend side human has been the trespass has been moved, it has to, uh, that.

Am I? Am I answered your question?

Yes, Professor, you've provided a thorough and insightful response. Your explanation of the challenges in handling multiclass data from sensors, videos, and images and the layered approach to decision-making using neural networks is clear and detailed. It shows a comprehensive understanding of implementing AI in real-world scenarios, particularly autonomous systems. I believe we’ve covered a lot today. Is there anything more you would like to discuss or highlight before we wrap this up?

How many minutes I am having, Sir?

We have just a few minutes remaining, Professor. If there's anything specific you'd like to emphasize or discuss further, feel free to share it now.

Uh, I would like to thank uh, VAT for, uh, offering me this uh, interview.

Uh, postings. And, uh, I am sure that I will be.

Prominent to work with the VAT with various designations and I saw that VIT have a large amount of scope in research. I will try to set up a few laboratories and other research areas of my interest and.

Thank

VAT's interest?

Thank you, Professor, for your thoughtful words and for sharing your aspirations. It has been a pleasure discussing your experiences and expertise. Shall we conclude our session here?

Anything I can add more to my this interview session?

Professor, you've covered your teaching methodology, research contributions, and approaches to advanced topics comprehensively. If you'd like, you could briefly highlight any specific plans or strategies you have for mentoring students or fostering interdisciplinary research at VIT. This might emphasize how you could contribute to the institution's growth.

Uh, thanks for the question. So I am, uh, mentoring students for a longer, larger period of time and most of the alumni is also still in contact. Uh, I try to focus on the students who are uh, uh, bright students has to be getting into the next level of their career and the students who are in the weaker section or middle level has to come to the next brighter levels in their careers. So.

I will uh, categorize the uh, we cannot uh, particularly categorize based on the skills we can give them, the more emphasizes in training and mentoring.

Uh, so to contribute this is to contribute towards the student mentoring, uh, and to contribute to VAT specifically in the area of research. Uh, I can uh, have a more number of collaborative research work with internal faculty and with external faculty or external research.

To set up a lab to create a new infrastructure, or else, uh, to create a.

Laboratories where Tesla and Google labs are like that working for a vehicle based.

Laboratories. So it will be helpful and grateful for me, uh, if I join VAT, we will be creating a more, uh, multidisciplinary, interdisciplinary and trans disciplinary research on vehicle and AI and marching towards the another new era of computing technologies.

Thank you, Professor, for sharing your vision. Your commitment to mentoring students at various levels and your focus on fostering interdisciplinary research are commendable. Creating dedicated labs and collaborative environments, as you've described, aligns well with advancing research frontiers. If you’re satisfied with everything discussed so far, shall we conclude the session?

Yes, thank you for your uh, interest and questions, what you have asked me for throughout this session.