OK, so quick introduction about myself. Thank you. First of all, thank you for giving me this opportunity and. My name is Pratyusha. And, umm. I just finished my PhD in 2025. At the moment I'm working as UMM. Postdoctoral researcher at University College Cork, Ireland. Uh, my research, umm, mainly focus on. Artificial intelligence during my PhD. I worked on securing personal voice assistants like. Google Home or Alexa? Umm, 3D devices. I specifically looked at, uh, protecting these devices from denial of service attacks. And. At the moment during in my. Postdoctoral research I'm working on. Uh, building LLM agents. To support the users. Umm for data monitoring. Ization especially. For protecting the data fairness and data governance tasks. So that's that's my current research right now. And prior to this I worked. In industry. Uh, while I was in India, I worked as an SAP consultant. Umm, where I worked on. You know SAP security and basis related. Roles. Umm. Apart from this, I also have been teaching in Griffith College Court from the past four years. And my teaching experience. Umm. Mostly includes teaching modules, uh, related to computer science. For example modules like relational databases. And web development, server side, web development, cybersecurity related modules. And. Also teaching modules like a business analytics using uh tools like Power BI and Tableau. Um, so this is my teaching experience as well, so. I would say that. I have a combination of. Uh, industrial experience with academic experience as well as the research experience. Altogether.

Please start by introducing your professional journey in academia

Uh, yeah. So.

Uh, right. So my professional journey in academia started as a PhD researcher at University College Cork, I mean. Even before that I was doing masters by research where I was, I just got, I just was introduced to research, uh, but I did not. Uh, fully work on it in the research perspective, so. As a researcher, my journey started. Uh, in CC? And, umm. As a as a researcher, uh, I. Basically worked on a lot of projects and I collaborated with the other researchers as well to publish the to publish some, uh, papers. And apart from that, like I said, I also do teaching. Before uh. Starting as a full time lecturer, I used to work as a. Uh tutor, uh, at uh, different college uh. Where I used to help students with UH assignments and. If for example, if they don't understand a certain concept in the class, I used to kind of, uh, teach them. In the tutorial, so that's what I was doing before. Working as a full time lecturer. So and then, uh, while I was doing my PHDI started. Teaching as well as a lecturer at Griffith College. Where I deliver full length modules. Not just delivering the modules, I'm also involved in activities like curriculum development for example, For example course trees, designing and also assignment creations. Exam paper writing. Marking the assignments and also, uh, managing students attendance and all these other administration activities related to the students as well. So yeah, my academic journey. Uh, started. With my BH. HD at UCC. And. As a researcher I was involved in. Umm, a lot of outreach activities, research activities, paper writing as well as teaching.

Just a quick intro — your name and what you've been working on lately.

Okay, you mentioned handling both curriculum development and student assessments along with research and outreach, which is a lot to balance. What would you say was the most challenging aspect of managing those multiple roles at the same time?

Yup. Umm. It definitely the. Most challenging part was dealing with students I. See, because students come from a lot of different backgrounds, lot of different learning abilities. And it wasn't easy for me to. Uh, initially wasn't that easy, uh, to kind of manage all the different, uh. Students with different learning abilities. So that's a kind of a challenge that I faced. The other activities like Outreach nor were part of my research anyway, so they were pretty straightforward and easier, but the challenging part was definitely managing people. Uh, in other terms, managing students. Of course it was challenging, but. Was I was able to kind of, uh, support them or help them, uh, you know, in ways I can so. Umm, later on I also learned. On how to deal with different UH students coming from different backgrounds. So the challenge basically turn into something. Very interesting. Learning experience for me.

You mentioned your PhD thesis on Personal Voice Assistant Security at University College Cork, and I see you published multiple papers on wake word jamming and adversarial training. How did you go about translating those research findings into concepts your undergraduate students could actually understand and apply in their coursework?

Yeah, so. Uh, my research. Mostly focused on artificial intelligence and, you know, some of the fancy terms that we hear these days. So if I have to break this down easily to students. Then I would definitely take examples. From the real world. Right, so my teaching is also a combination of umm. Taking the practical. Uh, experience from my, let's say for, for example, in the industry, whatever I've learned is. Whatever I've seen. That happens in the real world, right? So I try to bring that experience into my classroom as well. So if I have to explain. Certain concepts to my students, for example, uh. Voice Assistant, security, denial of service terms like this. I try to use real-world examples. So that they will understand it better. In other words, I try to kind of, uh, show them what, uh. Voice Assistant would look like I tried to. Practically. You know, talk to the Voice Assistant and tell them that this is where the attack can happen. This is how the attack can happen. If I have to talk about denial of. Uh, service. I will show them like. Uh, I can make this specific device unavailable for a moment by doing this certain attack on it, so. Any kind of a complex term or complex? Technology. If I have to teach such things, I would definitely use some real world experiences. And also, uh, from my industry experience, whatever I've seen. How the companies work in the real world. So I try to take that experience as well and try to teach them in that way so that they can. Uh, not just learn. By heart, but they will actually understand. What really happens out there, they can apply that knowledge. Uh, to something. More, you know, tangible. Yeah.

Right, so you actually bring your industry experience and real-world attacks into the classroom, not just theory. How do you handle students who struggle to connect those high-level artificial intelligence security concepts to the code or lab work—what do you do when the practical part just isn’t clicking for them?

Yes. That's for sure because students come from. A lot of different backgrounds. Not everyone will have the same learning abilities. Certain students will learn things much faster compared to others. I've also had experiences like this in my classroom as well. Where UH students, certain students would not understand. Uh, sometimes, you know, especially a very high level concepts. So if I. Have to explain it to them. So I do have a dedicated sessions. Basically. For example there is a theory session or even if it's not just. Uh module where there is a lab as well as a theory, even if there is only a theory session. So I do the teaching. And in between I kind of use tasks. So that you know that would grab their attention. So these tasks can include things like questioning them. So what would you do if? You kind of get into something like this, so. Things like that. So I would make them think about the scenario and I would make them apply. Or put themselves into that scenario to be able to answer these questions. And even then, if certain students don't understand. The the. Strategy that I use is I try to go to them. And I tried to sit with them and ask them. What is the problem? And if I can, I would definitely clearly break break the. Big concept into a smaller, uh, concepts and I try to explain them. Using those so that that would that had been always helpful for me. Breaking down a very complex concept into something. Much easier for them to understand. So that had actually helped students so far have seen a positive impact. Of teaching them this way by going to them and sitting with them and try to explain them. You know, taking them step by step, this has actually have seen a positive impact by doing this.

Okay, so you break down the tough concepts into smaller parts and work with students one-on-one if they're stuck—that's good to hear. When you build assignments or lab activities based on your research, how do you make sure they're fair and actually test student understanding, not just memorization?

Yeah. So I tried to create, uh, exercises in such a way that, uh, it would. Help them to learn it rather than. Uh, you know, just. Learning it by heart or, you know, mugging the things up. Of course that's not what we want. And of course in our computer science module, it's not that easy for them to just memorize certain things and apply them in. Uh, into the exam or a lab or, or a project. So that way it's already helpful. But. Even then, for example, if they are simply mugging up the steps or simply memorizing the steps. Umm, I would try to create. Exercises in the lab in such a way that will help them to kind of understand everything. Of how it's done, why it's done, rather than just, uh, you know. Memorizing and. Uh, and applying it on How so instead of just simply? Umm, answering the questions on like, you know what is it? How is it instead of that I try to create exercises to make them think? Why are we actually doing it? So I've been doing this during my. Umm, teaching over here, uh, in Griffith College. So that's how I. Try to create lab exercises. And also I would not expect them to. Kind of. Code everything by themselves. If, if at all, they're coding. Especially for such modules where there is a lot of coding involved, I also try to. Uh, code along with them. For example. During the theory session or during the demonstration session, I try to. Uh, do the coding part so that they see what I'm doing and sometimes I get into errors, right? So they also see how I am solving the error. So that they can also apply this. During the lab sessions as well. So I've been designing lab sessions in such a way. So that they can learn on. How and. Not just how along with how they also. Why they're actually doing it?

Dr. Pratyusha, looking at your current postdoctoral work on large language model agents for data governance and your earlier research on voice assistant security, where do you see the most fundable directions over the next couple of years? Are there specific grant opportunities or agencies you'd target to support this work?

OK, uh, so I believe that, uh. The most fundable direction in the few years would be definitely. In the speech domain, because it's a very. Uh, it's a. It's a domain that's nobody, really. Is looking into. And speech data actually contains a lot of parameters that can be used for a lot of reason. A lot of things, for example speech can be used as a. Medium to. Talk to the devices. To.

So you're seeing a lot of untapped potential in fundable speech-based research, especially with its application in user-device interaction. What's the current funding landscape for that area—are you seeing more success with government grants, EU calls, or industry partnerships?

I I would Bioma. So using. Speech Data. Would definitely uh. Would definitely have a lot of impact on especially. In the healthcare domain or you know, in other domains. So this is what I believe. So the most, uh, underexplored area is a speech. So if I have to find a fundable direction, so I would definitely use my experience on using the speech data for. Other, you know, applications such as healthcare and. For example. So this is where I wanna kind of explore.

So you're looking to leverage speech data especially for applications in healthcare and other domains that haven't really been explored yet. Can you give me a concrete example of a grant or funding body you'd target for that kind of work—and what you think would make your proposal stand out to reviewers?

Yeah, uh. It would stand out to the reviewers because, like I said, speech is a very underexplored. Area. Uh, that. Also contains a. A lot of. You know, insight on what all we can do with it. So. And especially if I want to write a proposal. Using this speech in the healthcare domain. Healthcare domain has a lot of. You know, impact in the society as well. So that is. A domain where? Definitely a lot of research has been going on. In that specific domain and, uh, considering its societal impact as well, I believe that, uh, if I write a proposal. Uh, in that specific. Area whereas such as using. Speech in the healthcare domain. I would definitely believe that. Reviewers would. Look at the impact that it can create. And. That way. Probably that would inspire them to give me the. Funding and the kind of grants or the kind of yeah, the kind of proposals I wanna write and the kind of grants I would wanna get is definitely the. Maddie, Curie. Fellowship, which has the. Call coming up in few more, uh, months. So I would definitely focus on that because even Mary Curie. Fellowship, Americori. Um, Grant is a very. You know, significant one and as a researcher. I would definitely want to kind of get, uh, some funding through it. So if I have to send this work off using speech signals. To understand the. You know. Physiological parameters of a human then I would definitely. At least send it to. Marie Curie Fellowship.

Alright, so you're targeting Marie Curie funding to explore speech signals for physiological assessment in healthcare, arguing that both the technical novelty and societal impact will stand out to reviewers. If you were building a new research group around this, how would you pitch its value to the university—beyond just the funding, what makes this a long-term asset for the institution?

Yeah, uh, this specific. Uh, domain, like I said, has a lot of. Tangible societal impact. And universities really should focus on. Uh, such kind of projects rather than getting a. Looking at the commercial aspect is what I believe, uh, because this specific uh project. Would definitely have a commercial aspect. For example, if we design an app and. You know, release it to the world and have industrial collaborations. That's clearly, uh. Uh, commercial gain. But. This specific project, like I said, also has a lot of societal impact. To it. So I would believe that if I build, if I try to create a research group around it, and I would definitely encourage them. Uh, to kind of, you know. So this specific project involves lots of. A collaboration with not just one specific domain, not just computer science, we also need collaborations with other healthcare. Domains as well. So thinking about that. This specific project. Which is being used by the university and collaborated with other hospitals and if it could really help patients, uh. Who are kind of. Uh, have who has like for example hypertension? A problem and if you could detect that with speech in a non invasive method. That's clearly a great impact it would have. Not just on the commercialization part, but also on the, uh, you know, helping the. Patients or you know the healthcare domain as well. So that way if I. See, the impact it has is not just limited to 1. Sector. It has impact on. A lot of uh sectors as well, so. I really believe that university would gain a lot, uh, not gain a lot. University would definitely have a lot of. Umm, you know, Yes, positive. Uh, impact on, uh, because of this specific project as well. So that's what I believe. It's not just the commercialization part, but university would. Definitely stand out. Among the other universities, I would say. If we kind of collaborate and do something like this. We would definitely stand out. Among others. Because of the impact that this specific project might have.

Alright, you’ve talked a lot about breaking down complex concepts and designing hands-on lab work. Suppose you’re teaching a class on artificial intelligence in Multimedia, but slides aren’t allowed and you’ve got two hundred students. How would you actually run a session that keeps everyone engaged and ensures the material sticks?

Feature of it is. I can actually talk to students and show them. How a machine learning model? Would change if I change the. Hypopitometers of it. So. Or if I I I can I also try to show them. I have a data set and if I want to predict something. I wanna play the accurate by looking at the accuracy, isn't it? So if I. Change the hyperparameters. How does the accuracy of the model changes or if the model is not at all suitable for the data set? I tried to choose a different umm. Model and that would kind of show them how by choosing a different model. How does the accuracy change? So the playground was a very interesting tool which had lots of visualizations in it. Uh, when I choose a certain model, uh, the visualization changes and the accuracy changes. The Roc curves and you know. All these things would change. The confusion matrix would change by looking at this, even though they did not probably follow my. Teaching throughout, but by talking about, by showing them the specific. Tool. Everybody's attention was on that specific tool when I was talking and showing them. The different models, confusion matrices and also the different hyperparameters. So whatever they've heard. If if also if they did not hear. My lecture. By talk showing the specific tool that definitely grabbed their attention to it and it also had a positive impact on the students that. They were kind of, uh. Then they had a lot of questions and it was it became a very interesting discussion. So if I have a classroom full of people, 200 people. And if I don't have. Proper presentation or slides to talk about the concept. I would definitely make use of available technologies. Uh, to make them understand the concept better.

You’ve clearly got experience innovating in your classroom and across research domains. Accreditation cycles can be demanding—if you joined our department and found that outcome assessment data was inconsistent across courses, how would you approach fixing that?

Outcome assessment data.

Sorry, I would like the question to be repeated one more time. Thank you.

Right, outcome assessment data is what accreditation bodies look at to verify student learning is being measured and reported accurately. If you noticed the data wasn’t consistent—say, some courses had detailed rubrics and others just had pass or fail marks—what would your first steps be to get everyone on the same page?

Sure, I'll repeat the question. Accreditation cycles can be demanding—if you joined our department and found that outcome assessment data was inconsistent across courses, how would you approach fixing that?

So. Trying to understand the question clearly. Outcome assessment. What would it mean by outcome assessment data?

You mentioned bringing practical industry experience into your teaching, but could you give me a concrete example of an actual industry project or consultancy you worked on—what was your role, and what specifically did you contribute?

OK.

Right, outcome assessment data is what accreditation bodies look at to verify student learning is being measured and reported accurately. If you noticed the data wasn’t consistent—say, some courses had different reporting formats or unclear criteria—how would you approach fixing that?

Uh, try to perform the internal modulation, uh, where applicable. In order to kind of. Make them consistent across the different courses. Yeah. So I would definitely. Definitely use. Internal moderation, etcetera. To kind of. Get this consistent across all the different courses, yeah.

Alright, so you'd use internal moderation to make the assessment consistent. Can you walk me through a time you've actually participated in moderation or standardization for assessment—what was your specific role and what did you do?

Yeah. Uh, we did have, uh, this kind of inconsistencies, uh. Not in my module, but I have seen other lecturers kind of. UMM went through this. So. For for some of them. For one of them, actually, I was being. I was chosen as an internal moderator. Where we try to kind of. Take the course objectives and we. Try to kind of compare it with what's being delivered and uh, we try to see. Whether the learning outcomes of the student really match with the object is we set initially. So. This is the kind of moderation we did. And yeah, so sometimes, most of the times this this issues will be fixed in the moderation stage, but if not, sometimes the lecture also will have to. Uh, attend the academic. Meetings with, uh, the, you know, like program share and, you know, head of school, all these things. So so this happened. And I've I mean. Even though I don't have a straight. Forward experience in this. I did see others kind of facing this and. Fixing out using the internal moderation steps by. Comparing the outcomes and the. Learning objectives, you know.

Okay, so you participated as an internal moderator comparing course objectives with delivered outcomes, and you saw these inconsistencies being discussed in academic meetings. I want to shift gears slightly—can you give me a concrete example of an actual industry project or consultancy you worked on, what your role was, and what kind of impact it had?

OK, so. I worked uh, so my professional journey started with. Uh, industry. So I worked. Uh, in a in various. Projects in Cognizant, that's where I worked in the industry. So I have worked on different projects and especially projects that. Our dealing with, you know, like the retail pharmaceutical that kind of. Client. Those were those were the clients. So. Of course I. Participated in a lot of things, so I also worked as a train. I also used to train people who were the entry level. Uh, engineers to the project, so I used to also train them. Umm, on, you know the. Tasks that they are supposed to do. So I was working as like a shift lead. Team lead for people of seven. And yeah, so the outcome that it had was. The team was successful. We were able to deliver the project on time. And uh. Yeah, it was. It was going quite well. So I do have a straight. Uh, industry experience. And like I said, I used to work on. SAP domain. With various retail projects and. Apart from that, I also, uh, acted as a shift lead. To entry level train. To train entry level. Engineers. Yeah.

Okay, so you worked on SAP projects in the retail pharmaceutical space at Cognizant, acted as a shift lead, and trained new engineers. When you were leading that team, did you ever have a project where something went off the rails—like a late delivery or a major technical issue—and how did you handle it?