Good day, Abeyram M Nithin. 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.

Yeah, hi. Is it audible?

Hello.

Hello.

Yeah, OK. Thank you for the confirmation. Yeah, myself Doctor Abhiram like I completed maybe B from government College of engineering in the stream of mechanical engineering in in the year two 2014. Then I did my postgraduate in CAD Cam from Mefco slang engineering College Sivagasi and I graduated there in the year 2016.

From there I did my.

Doctoral degree from NIT Warangal under the guidance of doctor MJ Davidson and Doctor CSP Rao.

And in the in the Department of Mechanical Engineering, my area of expertise is.

Extrusion like extrusion of powdered specimens and in this PhD work also I have used several computational tools like Deform 3D software and.

And, uh, I have even used machine learning, uh, uh, sorry, uh, deep learning algorithms in, uh, in my research work.

And I graduated from NAD Warangal in the year 2000 2020 2 Now I have been working as Assistant Professor for the past three years at Trinidad Institute of Science and Technology.

That's it, yeah.

Professor Nithin, let us begin by discussing your expertise in computational modeling, specifically in relation to its application in your research or teaching efforts. Could you outline a comprehensive example where you've utilized computational modeling effectively?

I, I started my computational work in my postgraduate which was in the year 2015. At that time I was involved in FEA simulation, finite element analysis simulations in which I was modeling A friction friction plug welding process. So that was like more very simplistic analysis which was carried out in ANSYS software. And then during my PhD I like. One of I I have published 4 papers in my from my PhD work.

Yeah, sure. I can do that. See, like at the very beginning, I started my computational work.

In that one paper was like related to even even that was uh, finite element analysis of extrusion process. So that was carried out using a deformed software, which is a metal forming specific software. So in that I have worked on dynamic recrystallization studies, which is like analyzing the microstructure of the, uh, extruded samples during extrusion process. So I have. Converted that work and.

Like I have published a paper on that work in the journal of in a SCA journal and at the towards the end of my PhD work I did. Which means the final objective. My last objective was about predicting the faces like faces in the microstructure images using deep learning algorithm. So there I have used.

Like residual alert, the residual net unit architectures which were used to predict the like say colored regions like say some area of interest would be there in the microstructure that I was predicting using deep learning algorithms.

And last two years I have been working on uh, like I have published 2 reputed journals in UH-1 in material characterization and another uh, and another in International Journal of Advanced Manufacturing process. So in this I have used machine vision techniques for analyzing the, uh, microstructure images and like classifying them. So these works I have done along with that I have also carried out.

Regression analysis on my, uh, flow stress behavior datasets, uh, uh, like and that has been published in a, a submit journal where like regression techniques has been used for predicting the stress from the input variables like strain temperature and strain rate. Uh, so this works have been done and like ongoing works also have been related to machine learning problems.

It is evident that your work spans a diverse range of computational techniques. I’d like you to reflect a bit deeper. In your recent works involving machine vision and regression analysis, how did you validate the computational models you developed to ensure their reliability for future applications?

So generally.

Like, uh, I will start with the regression problem. So in regression problems like how we have, uh, validated these things is on the basis of performance metrics like R square values, mean squared error values and RMSE values. On these basis only we have like the higher the R square values, which means the model is more perfectly fitting the ideal cases. So this is how we, we have validated, whereas in computer vision algorithms, there is no.

Like, uh, quantitative analysis only, uh, so there is no like metrics or one measure by which we will be evaluating that.

Metrics or something to validate that. Generally it'll be like, uh.

Your approach to validation through statistical metrics for regression problems is clear. However, regarding the computer vision models, could you clarify how you ensured their accuracy or robustness despite the absence of standardized metrics? For instance, did you employ confusion matrices, cross-validation, or other indirect methods to assess the classification tasks?

Uh, no this like uh K fold, K fold ma. See in this machine mission problems what we have done is.

Yeah, Yeah, yeah. OK. Our idea was to identify the, uh, grains in the microstructure. So those, yeah, those grains in the microstructure would be colored and identified as areas of interest. Then based on that, what we do is we used to measure the grain size. So like, uh, grain size means the number of grains in the, uh, in the particular image. So the those grain sizes would be measured using.

Uh, like our idea is not classification, whatever performance metrics you're saying, all these problems are related to classification. Whereas in mission mission problems, which I used, our idea was to identify, uh, like, uh.

Umm computer machine mission techniques and that could that would be compared with the umm grains is measurement which is manually done so on those basis this. This mission mission technique.

Grain size analysis has been validated manually.

Understood. So, your validation in the machine vision tasks relied on manual comparison of grain size measurements to confirm the effectiveness of your modeling methods. Let’s move forward—how have you integrated your computational and machine learning expertise into teaching? Specifically, what methodologies or techniques do you use to simplify these complex concepts for students?

Umm, see. Like, uh.

There are two approach approaches towards teaching any technology to students. It would be like see the the academic ways in academics, what we do is we just probably used to teach them the basics. Like more probably it would be the theoretical background. Hardly we used to cover any practical things behind that how the machine learning works. So by doing or by practice, you know there is like less scope in academics. Mostly academics deals with just like.

How the algorithm works or the behind the the mathematics behind it. Whereas how I approach my approach towards problems would be like I take a problem, I identify what are the like problems in it and what are the validation methods like how to finalize it and finish it. So based on that I start by trial and error method. So this is how I have been working. So as far as teaching is concerned, what we have to do is we have to blend both the things. Theoretical knowledge also should be provided parallely. Hands on expertise for students also should be.

Dot like say the a simple regression problem can be taught to them like we can start it with linear regression where students will come to know manually they can calculate the mean squared error and this R-squared values simultaneously. A small code, a code snippet using linear regression which is which probably would have been written in Python code or something could have been used and this no. This would help the students understand and familiarize with the concepts as well as they will get.

Hands on practical knowledge on that also. So that is how we can integrate both the practical knowledge as well as the theoretical knowledge into it.

Blending theoretical understanding with hands-on applications seems to be a well-structured approach. Starting with linear regression and providing both mathematical and coding insights offers students a strong foundation. Could you provide an example of a course or lab session where you implemented this blended teaching method effectively? How did the students respond?

So like the see this can be like say since I am from manufacturing background say generally manufacturing subjects would be there for Mechanical Engineers like mechanic, engine, btech, mechanical engineering student.

So in that case, what we can do is we can take a small data set like say of 27 experiments where probably by the end of second year most Mechanical Engineers like would be aware of the process parameters which are involved in machining process say like speed, speed, depth of cut and say we can use this speed, depth of cut as input parameters and probably like. Surface roughness or something as output parameter and try to.

Provide a very simple. This simplistic problem can be explained to the students probably by coding as well as by hands on calculation also. So that is how we can teach this concept to the students even during live sessions or like during classrooms also in classrooms also this can be taught to them.

A pragmatic approach indeed—leveraging a machining process dataset that students can relate to makes it much more tangible. By blending live coding sessions and manual calculations, students gain a clear understanding of both concepts and applications. Moving ahead, you’ve mentioned student evaluation in your credentials. Can you elaborate on how you assess students’ grasp of such interdisciplinary topics, especially when blending theory and practice?

So evaluating students like.

With the advent of like this a generative a, the need for learning coding has shrunk dramatically. There is no like if we if we know the basics that even that would be enough. You would no need to go to the advanced level of coding or to the intermediate level of coding, even basic level of coding.

With little knowledge and debugging would be enough. So to evaluate students, what we have to do is first we have to teach them the simple level of coding as well as the theoretical concepts say like take the example of machining. So what is speed, depth of cut and how it relates with say.

Surface roughness. So these things, this domain knowledge plus as a minimum amount of coding knowledge would be enough. This can be put to the test.

Uh either cutting force or average force or like some other output parameters can be given and we can ask the students to predict the same input parameter like speed feed, depth of cut as input and cutting forces the output. And we can ask the students to use linear regression to predict this and whether it is like. Whether it is in whether the derived results or in acceptance with the physics of the problem. So by this we can evaluate whether they have understood the physics of the problem.

Two students say by like just slightly changing the data set. Not only data set, some other parameters like say probably speed, feed depth of cut would be input parameters. Instead of surface roughness we can give like um.

As well as their the coding knowledge also.

This dual-layer evaluation method—assessing both the students’ grasp of the domain's fundamental physics and their application of computational techniques—seems robust and insightful. It tests their understanding holistically. Moving on, your resume mentions a Ph.D. in Mechanical Engineering along with research in materials characterization. How do you ensure a balance between your research commitments and teaching, particularly when both demand significant attention?

So, umm, this, this has been there in every institution whatever. If we are in the field of academia, we have to balance both the things say like teaching, uh, teaching students as well as doing research. I don't think it will be much difficult balancing this. Anyhow, we have to balance all these things. So, uh, I believe I can balance it more properly. I have been teaching for the last three years, so I don't think there will be any problem. The more the students are like.

Brilliant. It will be easy for me to teach as well as make the students work on projects also. So VAT being one of the top most institutions, the students calibre is also so good. So it will be not like you are pushing too hard. If you explain the concept clearly, the students will be able to understand these things as well as probably research work also will be. Can be easily navigated. I don't think there will be.

Be much problem in that.

Understood, Professor Nithin. It’s clear you are confident in managing both teaching and research. Lastly, you’ve highlighted consultancy and industry interactions in your achievements. Could you provide an example of how such industry involvement has informed or enriched either your teaching or research methodologies?

So right now the problem in academia is like we have been doing lot of research works but nothing is getting but but most of the works are not turning into any products be it hardware products or digital software like hardware products or software products. Only when this turns into some products there will be this cycle of funding and continue to doing research and that research again leading to a product that product further yielding money.

So if we want to get into this cycle, we have to do consultancy projects or in be in touch with industry so that whatever work we do will be related to the industry or we work on industrial problems. Either of the way either our work should be related to the industry or we have to get work from industry and solve the problem like whatever problems they have or issues are there, we have to address that and make a product out of it. Or solve their problems.

So that in in return, they will also keep funding the research and this cycle keeps going. So that will be a sustainable solution both for academia as well as.

It will be sustainable otherwise like we will be working in parallel like both industry as well as academia will be working in parallel words where both of both of them will not be able to meet their or achieve their end goals.

Industry The big problem in today's education system is there is a significant gap between academia and industry. So only when we close this gap by being in touch with industry.

Your emphasis on closing the gap between academia and industry is astute, particularly the idea of aligning academic research with industry problems to ensure sustainability. Professor, let's close by discussing your vision: how do you see yourself contributing, both in teaching and research, to a dynamic institution like this one in the next five years?

So actually my aim is to.

Head the center of excellence wherever I am like center of excellence on like interdisciplinary center of excellence on manufacturing.

Uh, finite element, like, uh, manufacturing, machine learning as well as finite element analysis. So what we do here is it's an interdisciplinary center where we work on manufacturing problems. Those problems will be simulated in like, uh, using softwares like ANSYS, Abacus or FEA softwares.

Datasets from that along with that we will be using.

Machine learning to process those datasets and.

Process those data assets and extract insights from this. So I want to head the center of excellence where we'll be working on hardware problems, plus work analyzing those things in the simulated environments. Along with that, machine learning work also will be integrated into this work.

Your interdisciplinary vision for a Center of Excellence that integrates manufacturing, finite element analysis, and machine learning is both ambitious and forward-looking. It demonstrates a passion for addressing complex real-world problems through academic rigor and innovation. Professor Nithin, this concludes the technical verification phase of our discussion. Would you like to proceed further or request to end the interview at this point?

Can you repeat it? I didn't hear it properly.

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

Certainly, Professor Nithin. This concludes the technical verification phase of our discussion. Would you like to proceed further, or do you wish to request to end the interview at this point?

I would like to proceed further.

Professor Nithin, let us delve into your academic research background. You have two notable projects listed on your resume. First, the development of "Constitutive and Machine Learning Based Microstructure and Mechanical Properties Prediction Models for Thixo Extruded Al/Si Alloys," and second, "Friction Taper Plug Welding for Repairing Operations." Starting with the microstructure prediction models project: Could you elaborate on your primary objectives in this research and the practical implications of your findings?

So, uh, this was my PhD project like, which I did for, uh, around six years. So in this work what I did was like, umm.

The first objective was about we take powders, aluminum powders, silicon powders, we mix it and we create powder compounds, powder compacts means like powder pellets. So these pellets, these compacted pellets has been prepared using hot compaction process. So this hot compacted specimens.

Then will be extruded in the extrusion dies the temperature range for the extrusion is like in the semisolid range. So what we do is we used to find the semi solid range of the particular aluminum composition and we used to process that or extrude that powdered pellets at that extrusion ranges. So this was my overall work so this.

Powder compacts as well as extruded specimens. Characterization has been done.

So this has been done characterization tests like density, hardness.

Density, hardness and stress strain plots has been.

Taken and analysis has been done so in characterization density.

Uh, yeah. Uh, density and hardness has been studied, whereas in characterization, what you know in stress strain constitute analysis. What we did was we. Uh, Cal, we found the process parameters for this Johnson Johnson Cook model and we have analyzed that along with that I have done.

Mission like deep learning. I have used deep learning algorithms for predicting the phases which are present in the microstructure.

So this is the work which I have done in my PhD, whereas in my.

That was a friction taper plug welding. So this the idea of this work is to if there are cracks in the plates or like aerospace structures, what we used to do is for repairing what we used to do is we used to create a hole in that and a taper will be.

PG. What I did was.

Uh, filtered on that cracked region and it will be fitted using friction welding process so that it becomes a solid-state joint instead of a, a heat based joint. So this is what we did there. And yeah, uh, this was second PG project was a more small project compared in comparison with the. PhD project. So here what we did was we carried out.

Yeah. Well, in this, what we did was we carried out characterization test like bending test.

Microstructure analysis, hardness test and these tests were done and the works were completed for this PG project.

You have provided a thorough overview of your research. I would like to follow up on two specific aspects. First, regarding your PhD project—predictive modeling using deep learning algorithms for microstructure phase prediction. Can you discuss how you ensured accuracy in your predictive models, especially given the variability inherent in microstructure compositions? Second, transitioning to your PG project—Friction Taper Plug Welding. From a pedagogical perspective, how would you simplify the explanation of this innovative repairing technique to undergraduate students or those unfamiliar with advanced welding methods?

Our scheduled time for the interview is now up. Before we conclude, do you have any questions related to the role or the company?

Uh, like can I know the interview pattern? How it will be the physical interview pattern?

Like will it be, uh, the what? What topics will be rela what? What kind of questions can I expect in the interview?

In the physical interview.

I cannot provide specifics about the interview pattern or the questions you might encounter in the physical round. However, you can typically expect discussions centered around your research, pedagogy, and contributions to the academic community, as well as potential hypothetical or real-world applications of your expertise. If you have further concerns about preparation, I'd recommend consulting the department directly for any additional insights. Was there anything else you'd like to ask?

That's it. That's it for now. Thank you. Thank you very much.

No.

Thank you, Professor Nithin. Have a great day and all the best in your pursuits. Take care!

Yeah. Thank you.

You're welcome. Could you elaborate on your 3-year research roadmap, specifically focusing on how your work in computational modeling will contribute to institutional rankings and revenue generation?

No, like in the last three years I have published around six journal papers.

Out of which?

Like out of which three-year SCAQ and journals?

And like rest of the papers has been, one paper has been.

CA Q2 with other journals be with other two papers being SCA Q2 papers.

Thank you for detailing your publication history, Professor. Could you explain how you plan to scale the efforts of your lab over the next three years? Specifically, how would you mentor junior researchers to ensure a steady pipeline of high-quality publications?

So, like, uh, uh, mentoring students, uh.

Like mentoring student, uh, mentoring students will be will need a small at least some kind of umm.

So the students being exposed to some kind of, uh, like, uh, miss, like computer science courses related to machine learning or like small value-added courses, uh, or short-term training programs for students so that they get to know the basics of, uh, machine learning. Where, because nowadays this knowledge of a is a must for every student, every researcher, every faculty, irrespective of their domain, because like we are.

To deal with some kind of data sets or something and we need to work on that. So a minimum like say probably minimum computational programming knowledge is a must for everyone. So that should be brought into the curriculum so that.

So that like guiding the students or like motivated, like teaching these skills should not be an extra should not consume extra time if they consume time in.

Along with the along with the curriculum itself, we will have enough more time for the students to practice these techniques rather than. Until this computational skills are integrated into the curriculum we will be spending more time on teaching them this computational skills as well as the making practice them these things. So if we integrate them into the curriculum it will be more better so that they will.

Know the basics of these things at the first step. The second step would be like exposing them to problems, giving them some problems so that they work on that and they will be they will get enough practice on those problems.

And once they get probably these all are these basic levels of problems. So that would be enough at the UG level. So maybe for PG or PhD level students we may have to go to the intermediate level.

Of say like coding or like the complexity of the problems has to be slightly increased or intermediate level problems so that probably PG students and research scholars will be able to deal with much higher problems in their future.