 |  |
 |  |
Join us on LinkedIn Live for a compelling session with Swagata Ashwani, a renowned expert in data science and AI. This event, ‘The Data Alchemist: Transforming Numbers into Insights’, is a must-attend for anyone looking to master the art of data.
Swagata will guide you through the revolutionary world of data-driven decision-making, revealing how AI is reshaping healthcare, finance, and more.
Discover the importance of AI transparency in critical sectors and understand the urgent need for gender diversity in tech.
Learn how AI is becoming a formidable ally in cybersecurity and delve into the concept of Data as a Service.
Discuss ethical AI, the art of data storytelling, and the application of AI in emotional well-being.
This session is an invaluable journey into the heart of data science, offering insights that bridge the gap between theory and practice in the tech world. Enhance your skills and knowledge in the rapidly evolving field of AI and data analytics.
Follow Swagata on LinkedIn: https://www.linkedin.com/in/swagata-ashwani/
👉 See our Website: https://podcast.boomerliving.tv/
🎙 Boomer Living Podcast: https://hanhdbrown.com/
👉 LinkedIn: https://bit.ly/2TFIbbd
Transcript:
Hanh:
Welcome to Boomer Living Broadcast, where we explore the exciting blend of AI. Aging and the digital innovation. I’m Hanh Brown, and together with AI50 and Microsoft Startup, our goal is very simple, to make AI understandable and useful for everyone. Our approach is rooted in Azure’s technology, focusing on building a unified AI ecosystem that is
Hanh:
safe, private, and cost effective, particularly for senior care. Our journey isn’t just about cutting edge technology, it’s about creating a connected world where AI solutions are easy, safe, and designed with everyone in mind, especially for the aging population. So we’re dedicated to developing a unified AI platform that enhances life quality and bring us closer together.
Hanh:
So each week we dive into discussions with experts about how this AI. Unified AI ecosystem is transforming businesses and lives of older adults. So whether you’re a professional in the senior care, a caregiver, or just simply keen on the intersection, aging, health, and tech, this is a spot for you for engaging conversations and much insights. So join us as we pave the way
Hanh:
towards a more connected and compassionate era of senior care. So let’s get started. So today’s topic is data alchemists transforming numbers into insights. And our guest today is Skogata, Skogata Ashwani, the principal data scientist at Lume. She stands out with her expertise in health informatics.
Hanh:
and machine learning. A Carnegie Mellon graduate, she has a rich background in applying machine learning to mental health research. Her professional journey includes significant roles at Amazon and Deloitte, highlighting her diverse skills in data science and cyber risk management. She’s also an active contributor in the data science community.
Hanh:
She advocates for explainable AI and gender equality. Sharing her knowledge through various platforms and her leadership role at Women in Data. So her work exemplifies A commitment to innovation and making technology accessible and transparent. So Swagata, welcome to the show.
Swagata:
Thank you so much for the warm welcome.
Hanh:
Thank you. Thank you so much for being here. So can you share with us something about yourself? Professionally or personally?
Swagata:
Absolutely. So, like you said, my current role is at Boomi. I am working as a principal data scientist. And apart from my professional journey so far in data science, I actively collaborate and work towards women in data and those kind of initiatives that focus on bridging
Swagata:
the gap between like gender diversity. So I want more women to get involved in, in tech and educate them and mentor them. And a couple of years ago, when I was looking for a job, not just job, like getting more involved in the data field and AI field, I did not have that much of outreach. And I had to like literally reach out to people on LinkedIn.
Swagata:
So I wanted to create a platform where it’s easy for people to Especially women to come together, share their insecurities, share what they want to do in life, how they want to get there and create an ecosystem where they can learn and contribute and get involved in the field of AI. And I also, uh, like you mentioned in the introduction, I am very passionate about
Swagata:
explainable AI because as AI is advancing at the pace that it is, it’s extremely overwhelming and sometimes it gets difficult to understand what’s happening behind all of these complicated models. And, uh, so I wanted to. educate people on how they can use these tools, uh, whether they just want to understand why certain prediction was made or they want to just explore what’s
Swagata:
the possibility of what AI can do. These kind of tools can, can educate them and they can learn what their systems are doing in a better way.
Hanh:
Mm hmm. Bless you for that mission. And I echo that. I love what you’re doing. If there’s any way that we can add value, I’d love to. So, in the critical sectors like healthcare and finance, explainable AI is becoming incredibly important. So, can you discuss why transparency
Hanh:
in AI models is so crucial for accountability and trust in these fields?
Swagata:
Absolutely. Absolutely. So a couple of years ago, I was working for a healthcare insurance, health insurance company, and we were building a tool which would suggest where the patient should be discharged after a post acute condition. Now, again, this, when it comes to healthcare, machine learning can
Swagata:
only aid, it cannot dictate because Ultimately, it’s, it’s so important to make a decision with, with doctors and physicians who are experts. So, it’s only going to act as an aid, but with the current technology, I see that the, the predictions that are made are so detailed and, uh, we need to, we need to understand like why a certain prediction was made.
Swagata:
So, let’s say I’m giving a suggestion, hey, this, this person who has been, Uh, on rehab now needs to extend his, his or her rehab period. Um, there needs to be an explanation as to why that particular decision was made so that even the physician or doctors or whoever are validating that decision understand, okay, this is the reason. And yes, I think I agree or I don’t agree.
Swagata:
And for physicians to trust the system, trust the AI models, it’s important that it’s transparent so that they can trust it and then rely on it a little bit more than, than. the current situation.
Hanh:
I think what you said about it can be an aid, certainly not a replacement. Of course. Because these are data points that’s gathered and it’s from, you know, abundance of data and, and, and I think. The point, you know, how people say, you know, they’re worried about job replacement, which is very real, but I think more so, um, it is your companion.
Hanh:
It is your co pilot, not necessarily to replace you. So it’s all in your mindset and how you form your model and execute your prompts.
Swagata:
Absolutely. I agree. It’s not going to replace you. It’s only going to aid you and make you do like more creative and strategic tasks. So, enjoy. Enjoy this.
Hanh:
Yeah, no, I agree. I think it, it could make you to be more the improved version of yourself, you know?
Swagata:
Yeah, absolutely.
Hanh:
So, now, what challenges do developers face when creating transparent AI models? And how do you, or how do they overcome this?
Swagata:
So a lot of available tools that we have that are trying to solve the explainable AI, uh, they are focusing on, um, they’re called like post modeling phase. So let’s say you build a model. And now I’m trying to use a tool that will be able to explain that AI model. So as a developer, I am building something which will try to understand
Swagata:
something that already exists. And sometimes the explanation may not be a hundred percent accurate. Uh, whereas there are other kinds of research that is happening where while you are building the model itself. You try to understand it rather than doing the post analysis work. So if we, if we are able to design something which is a combination of
Swagata:
both, that will be very powerful. Right now, the biggest struggle is that, hey, we, we are providing this explanation, but we don’t know whether it’s 100 percent accurate or not. And that itself, uh, poses a big, big question mark whether or not should we trust this system, which it’s saying that it’s, it doesn’t know whether the explanation is valid or not.
Swagata:
But apart from that are other challenges. Convincing, convincing people that explanation is, is so important because a lot of times people just want to build something and they, they, they’re okay with that being a black box and, you know, as long as, as you get it done, that’s more than enough, but they don’t invest enough time in, in getting into the explainability aspect of it because
Swagata:
it’s still a nice to have feature, it’s not the most critical feature. So if you are able to tackle those two roadblocks, I think we have a, we’ll have a very good product in AI. And people will trust it more.
Hanh:
Yeah. Because when you say the word black box, well, you know, when you, when we’re trying to encourage businesses to use AI, we cannot work in a black box.
Swagata:
Yeah.
Hanh:
Right.
Swagata:
Yeah.
Hanh:
And then not to mention the, the key component in this is trust. And when you work in a vacuum. You do not earn trust. So it’s very important. And I keep thinking too, you know, we’re not in the business of building models. We’re in a business of solving problems to enhance productivity, efficiency, um, in the business,
Hanh:
but ultimately the consumers.
Swagata:
Absolutely. Yeah.
Hanh:
Yeah. So now, in your view, how can organizations balance the complexity of AI models, with the need for transparency to ensure trust?
Swagata:
I think, um, a lot of it will come with experimentation and showing the value of it. So for example, in my organization or organizations, I’ve been part of, uh, just presenting the value of it in terms of metrics and, and, and yeah, metrics mainly like. Hey, hypothetically, if we have explainability, what will your
Swagata:
tool look like and how it will set you apart from other people who are doing the same business? That is a key, uh, key metric to like force them or like try to trying to make them see the value of it. So that’s something that I’ve been trying to advocate. People that I work with that’s, Hey, if we incorporate this, this is how your
Swagata:
product will look like and how it’s going to add value to your product and how people will trust it more, use it more. And in the end, everybody’s a winner. So, so investing that little bit time and energy and putting resources into this research, it will go a long way. So I think that’s something that people who are trying to get into this field are trying to encourage their employers
Swagata:
can do something like pitch it, do some, run some simple experiments. There are open source tools that are available, which are very easy to use, play with it and just just show a demo and it will definitely convince anyone and everyone who are trying to get into it.
Hanh:
Mm hmm. You know, I agree. Even for myself, I think just like anything, power comes responsibility. Absolutely. And. Yeah. Many people are shy, reluctant, fearful, you know, for right reasons, but I always say start small and
Hanh:
gain confidence on those small wins. And then once you experience the wins and realize, well, gee, I just say X amount of hours or even tasks, you’ll be amazed.
Swagata:
Yeah. Yeah, yeah, I agree. Yeah. Small wins and like small, simple experiments can go a long way.
Hanh:
Yeah. And then you slowly. Take on bigger tasks to improve the efficiency and productivity and so forth. But, um, you know, for someone to come to, let’s say you, I said, you know, so how do I integrate AI? I always tell people, you know, I don’t know your business. I cannot tell you how to integrate,
Hanh:
but I would ask first that. It’s always best if you know what your pain points are, right? Because through those pain points, that’s where we can integrate AI, you know, steps at a time so that for you to gain confidence because, um, it’s, it’s, it’s not a silver bullet solution. It’s, it’s a enhancement augmentation, not a replacement, but you
Hanh:
still need to know your business and where you want to enhance. Yeah. It’s not going to do that for you if you don’t already know that.
Swagata:
Yeah, that’s where like domain knowledge is so important because if you don’t have domain knowledge, how are you going to use AI for your specific needs?
Hanh:
Right, right. And even a consultant, I got to tell you, you can’t. You can’t justify, explain, or sell the idea of AI. I think that’s very unique to the business. They have to go through that journey and come to a realization. You know, here are my pain points.
Hanh:
Yeah. And here’s how I need help in. And then that collaboration is much more fruitful. Because to tell a business who might think. I’m doing great. Yeah. Who needs this? And not to mention, I gotta go through
Hanh:
a learning curve and then one might think of all the possible reasons why it doesn’t work because that’s true. It may not work right away. There is a learning curve. So I guess what I’m saying is there’s a huge paradigm shift, uh, aside from the technology shift?
Swagata:
True, true. And sometimes I also think that everybody doesn’t need AI for their solution. It could be sometimes that you learn that what this technology does. And you can start small, it could be like a simple rule based thing that you develop first, and then you slowly iterate and add certain components of AI models. Most people try to jump into that
Swagata:
AI buzz because we need to do it because everybody else is doing it. But, but you can start small, see what fits you and then iteratively get involved more.
Hanh:
I echo that. If it doesn’t solve So if you have a problem of yours, it doesn’t do anything. It’s a fad and we know that it’s not. It’s here, it’s huge, it’s too big to ignore and it can be very integral to your day to day productively.
Swagata:
Yeah. I agree.
Hanh:
So now, you’ve experienced a progression from senior to principal roles. Within major tech firms. So can you share some insights on how this career evolution unfolded and the new challenges that it brings?
Swagata:
Yeah, absolutely. So I think the major aspect when you change roles from whatever position you are and get onto the next level, the key aspect I have observed at any organization is you already need to be performing at a senior level, if you’re a junior, for some time to be recognized and, and to get onto the next level. So for example, if you’re at a
Swagata:
senior level and you’re an individual contributor, you’re doing your day to day tasks, you’re, you’re reporting to your manager, that’s good. But if you want to get on to the next level, take a look at what are the requirements for the next level. Like, like let’s say for in my case, the additional responsibilities were. You also need to mentor junior
Swagata:
people in the team, guide them and, and, and work with them. What can they do better? So guiding, guiding and mentoring is a very important aspect as you go on to your next level. So start doing that and, and people will recognize that. And on top of that, you have to do certain things beyond your day to
Swagata:
day responsibility to be recognized. So be proactive. Let’s say if your organization is doing partnering with other teams on some research initiatives or some, some small. Experiments or POCs get involved and volunteer and say, Hey, I, I would like to get involved in this initiative. What can I do? So, and not just technical stuff, maybe
Swagata:
you can, you can volunteer to organize a team meeting or team gathering outside of your work, um, team building activities or doing things which are beyond your, you know, to do’s requirements. You start doing that and people will notice it. And then that’s how you, you get onto the next level. So in my case, I was doing that and I was,
Swagata:
I already started mentoring junior data scientists when I was at the senior level. And that just happened organically because we hired a lot of people and they were new to the company. So I just started mentoring them and then eventually, you know, without even realizing organically, you start doing. A little bit more and you enjoy the process, not just like looking at
Swagata:
it as a, as a career progression. You enjoy that process and yeah, that’s how it’s, it’s been great to be honest. Like I enjoy the balance of having a little bit of technical work at the same time, having some domain and leadership work as well at the same time, you know, uh, vouching for different tools and different initiatives like explainable AI and diversity.
Swagata:
So it’s, it’s a bit of very good, good balance.
Hanh:
Yeah. I can imagine. I think anyone would be grateful to have you as a mentor. And I think that’s one of the key attributes of leaders who are on the move is to create leaders. So, yeah, I think that’s wonderful. All right. So now what key skills and attributes.
Hanh:
Have you that you think it’s essential in advancing to higher roles in the tech industry?
Swagata:
In this day and age? I think having really good knowledge technically. Um, so we say that, hey, I will replace. All the jobs. Only a person who understands AI can use AI, right? So if you don’t understand what the code is doing or what your model is doing, you’re not going
Swagata:
to be able to make best use of it. So having good fundamental knowledge in, in data science or AI or whatever field you’re choosing. That is, again, very important. Even if you think that, Oh, I can, I can Google it, or I can, I can use your fundamentals have to be very strong for you to, to proceed in, in, in this technical field, apart from that,
Swagata:
I think good communication skills, it’s very important as you proceed in your, in your career, because only if you’re able to share your findings. Uh, to leadership or to management or convince that what you’ve done is, is good enough. That’s, that’s the key to, to have a good, uh, good progression in your career. So communication, good technical skills.
Swagata:
And also like we do something apart from, like I said, like do some mentoring, organize events, volunteer. Those kind of things really, uh. We’ll have a long term effect rather than immediate effect. So I think if you do these three, you, you are in a good position.
Hanh:
Mm hmm. Yeah, very true. So, can you discuss, let’s say, a particular challenge you face in an AI project and how did you address it?
Swagata:
Uh, yeah. So we’ve been trying to, um, build like something, uh, for our company, which is similar to what Most companies are doing like the chat GPT equivalent. So the main issue that most companies and us as well face is the privacy issue, because These AI models, as promising they sound, as good they are, they, it’s difficult for us to trust them, again, I
Swagata:
think because of we don’t want to share our customers or private data with them because we don’t understand, understand what’s happening in the black box. Uh, the data privacy issue has become a key challenge. So we’ve been looking for things where, hey, can we leverage these AI models and create something of our own? So that we are still protecting our data
Swagata:
and, and taking care of that privacy issues at our end rather than giving some third party the entire control. So the way we try to solve this privacy issue was, was using some fine tuning techniques. So we took the existing AI model and, and just fine tuned it in our own ecosystem so that we don’t, we have the, we have the power and we can control
Swagata:
the privacy and we don’t compromise on, on the privacy of our customers. So I think that is one of the biggest concern people have with these AI models. So fine tuning has become also a very good option that people can look into rather than giving that black box model control. You just take in, understand it, tweak it a little bit with your own data, and then you have the control
Swagata:
of your security infrastructure.
Hanh:
Wow. You hit it right on the dot because the fact that you said ecosystem, that’s huge. To me, ecosystem is having your own foundation model. That’s everything, right? Yeah. And that foundation model is when you take ownership of your data. And data is more than just
Hanh:
information, it’s your differentiator. That is what separates you from the rest because let’s face it, we all have the same tools. Yeah. We all have the same, but what makes us unique? That represents our brand, our voice, our customers, or the data. Yes.
Hanh:
Never give that away.
Swagata:
Yeah. Never give your data away.
Hanh:
Right. And that is like the underlying, um, understanding of having your own foundation model. And then under the entire ecosystems where you might not only have a chat assistance, you could have multitudes, whatever that might be over time, but the key is. You own the, the ecosystem driven by AI and I think until then you
Hanh:
truly reap the benefits of AI.
Swagata:
I agree completely. Yes.
Hanh:
Yeah. All right. So let’s move to ethics of machine learning. So ethical management of sensitive data is critical in machine learning, especially in fields like healthcare. So can you elaborate on the ethical consideration and practices, that should be put in place?
Swagata:
Yes. So, um, that’s very important. Like you have to ensure that you are making sure that you are like following all the regulations like healthcare. A lot of times we build like amazing tools, but we are not able to get into the deployment or the production level because of some regulations. So for example, healthcare
Swagata:
has HIPAA regulations. So Anytime you’re working in a critical industry, if it’s healthcare or finance or legal, first thing is to understand what the regulations are, what I need to do before I even plan out and start building anything technical, whether it’s feasible or not. A lot of times people do a lot of research and build great products, but
Swagata:
they’re not following the compliance or You are not following the regulations. So the first important thing is to make sure you are compliant. Make sure you follow the regulations. Talk to people if you’re not compliant and get it out of the way. That’s the first thing to do. And then proceed to building something amazing.
Swagata:
And even inside your data, you have such sensitive data. Make sure you’re following encryption policies or however, whatever the requirements are to make sure your data is secure. Uh, it’s encrypted in whatever format that industry needs. So just follow those protocols. It’s very easy to do, but most
Swagata:
people neglect it because they focus on building the product. But even if you build a product and it’s, it doesn’t get out of the window, there’s no use. So follow the regulations, follow the policies, follow protocols whenever you’re building in, in sensitive industries, especially.
Hanh:
And you know, I keep saying this, we’re not in the business of building models. We’re in a business of solving problems using AI.
Swagata:
Yeah. I like that a lot, actually.
Hanh:
And you know, to get the best adoption of your product, you have to have the end user forward feedback from the get go. Yeah. And that feedback, first and foremost, should include those guardrails or the policies and the HIPAA and so forth. So that’s how you gain adoption from the end users.
Hanh:
And, you know, again, we can’t work in a silo. Yeah. Okay. We’re not in a business of just building models. It’s about solving problems using AI. So I echo what you’re saying. Yes. Yes.
Hanh:
Now, safeguards and policies. So is that something driven by the industry, the client? Can you be more definitive on that? Because it sounds like it’s, there’s not a focal place that would define what those policies are. And it may be. And that is definitive enough.
Swagata:
So it really depends on, on, on company to company and industry to industry. So every industry like healthcare as such has a lot of government regulations and HIPAA. So that is the first layer of filtration you need to do. And the second layer of filtration is like if you’re working in
Swagata:
insurance, for example, again there are insurance specific protocols that you need to follow that the insurance industry as a whole follows. And the third layer is. the, the company wide protocol. So depending on your company, there are certain rules and regulations that the company has. For example, in our case, we, uh, as
Swagata:
part of my health insurance experience, uh, we had to follow like, first of all, the HIPAA and then, and then we had company specific rules that, hey, until and unless we get certain number of metrics, um, certain number of users. So we had a set of beta users who were nurses and physicians who used to test our tool unless and until they don’t find everything looks good, not just
Swagata:
in terms of model accuracy and those technical aspects, ensure that we are not giving out any personal information. Um, everything is, is in place. So those 200 set of beta users used to fact check everything in the, in the beta version. So there are three layers, I would say. First is the, the industry that you’re working on and then the sub industry,
Swagata:
like even inside health, there is, uh, insurance is a different category. So insurance has their own set of rules and protocols and guardrails. And then as a company, what have you set up? So for our case, we had like, hey, these 200 beta users have to approve it. Then only we will release it. So, those are the three layers that
Swagata:
I have worked on and that has worked really well in terms of ensuring that we are ethically grounded.
Hanh:
Mm hmm. Now, have you found a situation where, let’s say, your client may not have all those policies in place? And then how do you advise them? Where do you get it?
Swagata:
So a lot of small, smaller companies and startups, I think, have this issue where, where they don’t know. Um, a lot of companies in healthcare, they started off to solve the insurance problem and they build in great tools, but they were not compliant and they didn’t follow the protocols. And that’s, even though they had a great product, they were not able to get it out.
Swagata:
So I think, um, the best practice to do is getting in touch with, with these organizations. Who, who are in charge of making these protocols. So we had a dedicated team inside our organization who used to work with, with, with a compliance team who used to give us the set of, Hey, these are the compliance. These are the latest HIPAA practices.
Swagata:
And these are the things that you need to follow in your product. So, we had a dedicated team who got in touch with, with the, with whoever is writing these protocols and getting in touch with the government organizations and making sure they are up to date and they know what is the latest change and they would just share like bullets with us.
Swagata:
Hey, these are the things you need to follow. And that kind of, because we were a bigger organization, we had that advantage that we have a dedicated team. But if you’re a smaller company and you don’t know. Get in touch with, with these organizations and, and get it out of the window.
Swagata:
First things before you start even building your product.
Hanh:
So true. So now, how do you foresee the role of ethics evolving in machine learning and data science in the near future?
Swagata:
It’s going to be huge. I think now ethics will comprise of a lot of aspects like responsible AI, explainable AI. Privacy, security, compliance. So I think ethics initially was just like being compliant. Now it’s going to expand to these newer terms like, you know, responsible AI, where you’re making sure that,
Swagata:
that your models are not having any, any kind of bias or, uh, doing some sort of discrimination because of historically how it’s trained on it. I’ve seen examples where, you know, you have these models who are suggesting job and they are biased towards men. So these kinds of things, uh, happen a lot because of the data and our data is biased.
Swagata:
I mean, historically there has been a bias. So how do you ensure that you are making sure you put in the right, uh, techniques in order to make your data more, more balanced? So I think ethical ethics in machine learning and AI is going to expand to a lot more than what it is right now. It’s going to expand to responsible AI,
Swagata:
explainable AI, security, privacy, bias. It’s, it’s going to become a big part of the machine learning process.
Hanh:
I agree. And then as AI advances. All those key elements that you describe are going to be expedient, exponentially huge. And I think, um, it’s good for the consumers. It’s good for business. You know, it holds us responsible. I think it’s a win win if
Hanh:
we all can do this right. And I, I always go back about the ethics, you know, for instance, in my mind, your prompts, your directives will lead you to an output in using AI and ask yourself, do these outputs impose anything negative on your family and your children, let’s say, and if your answer is yes, or, you know, you got to rethink. So, put it in a personal perspective.
Hanh:
Absolutely. So, that’s how you determine if it’s ethical or not. I mean.
Swagata:
Yes, I agree. Yeah. I do that a lot. Like, you know, try to think about it personally. Like, would you do it if you were the CEO or, yeah, that, that makes you, you know, shift your moral compass and, and do the right thing.
Hanh:
Mm hmm. It’s AI empathy. And it’s, um, it’s not technology. It’s we directing the output. So, um, all right. So now your work in using machine learning to understand college students emotions is very innovative. So now can you share what you’ve discovered and how this can help beyond
Hanh:
just students, for instance, older adults?
Swagata:
Absolutely. So I think, um, what, what I did was as part of my research, I was looking into sophomore students. And we luckily got access to some students who volunteered to share their, their text messages, um, Facebook and, and just iMessages for a period of three weeks. And we tried to analyze and their smartphone usage as well.
Swagata:
So using that data was very helpful to understand. Who in that group is more prone to depression or mental illness and uh, interesting find was that there was a very direct relationship between what they post on, on their social media and how much time they use their social media. So people who use. A lot of social media and, um, post a
Swagata:
lot of content related to, you know, what they like on social media, like, for example, on Facebook, if you’re liking content related to being alone, of course, that’s a very naive example, but things like that are very naive. strong indicators that, that you are not feeling well and you could be depressed in the future. So we, we use that, that learning
Swagata:
and kind of build a tool as a, as a virtual assistant or, uh, which could help you help you talk to somebody and get help if you need it. So the way that tool worked was that nobody wants to, first of all, acknowledge that, that most people do not want to acknowledge that they are facing depression.
Swagata:
So that’s the first, the first thing is to make that tool very interactive and very easy to use and very friendly. So it’s like a, it’s like AI buddy for you. And once you get friendly with it, that’s when that tool takes you and asks you, Hey, is the fun, there’s a fun therapy session. How can you make therapy sound fun?
Swagata:
Uh, so that tool kind of asks, becomes your friend first, and then navigates you to this, this. This therapy that you might need. So identifying students who are, uh, who could be prone to depression, they are given access to this tool and then they can use it for their advantage. So similarly, I think if you want to draw this parallel to, to older adults,
Swagata:
um, I don’t know if they have that much smartphone use phones or they use it or not, but they have to figure out some other, um, Other metrics that we can use and we can get their data from, like, what do they do during the day if we have some, some way of accessing that data, what they do, do on a daily basis. and identify the, the key aspects of how we can help them out and find out
Swagata:
people who are prone to depression or mental health issues and then use the same tool and kind of help them out.
Hanh:
Mm hmm. Mm hmm. No, that’s great. I think if we can somehow extrapolate that study outside or beyond of students to older adults, aging population, it’s much needed because I think they are the ones suffering from loneliness the most.
Swagata:
Absolutely.
Hanh:
Very crucial work. Now, so how does machine learning help in getting a clear picture of someone’s? Emotional health and like what methods did you find most useful?
Swagata:
So first thing is that, um, need to have good domain knowledge. So in our case, we knew that we are targeting teenagers and teenagers spend a lot of time on their smartphones and, and what social media apps they use, uh, where they hang out, their location. Those things were key factors in understanding the domain. And then the second aspect was another
Swagata:
interesting aspect we figured out. Was that whether they are in a relationship or not, because that’s also a key indicator of whether you are feeling lonely or have any mental health issues in terms of building machine learning models. That is the easiest part. I would say, like, once you have good data, once you have good understanding of your of your problem and the
Swagata:
solution that you’re targeting, building model is pretty straightforward. Like we, we experimented with bunch of popular, uh, machine learning models back. It was, this was five years ago. So we didn’t have like very sophisticated models that we have right now, but all the models that we experimented, the classical machine learning models. they gave us pretty good
Swagata:
understanding of why certain people. So once we got a prediction, then we back traced and looked at, okay, these were the features of that person who is suggested as being depressed in the future. And we took a look at what are the common aspects that we can find in all the people that, that were. Predicted to have depression and we could really get really good insights like,
Swagata:
hey, okay, not being in a relationship is one, um, using social media at night at, at those, you know, 3 a. m. and those kind of times is another indicator and liking posts, which are like. You know, very weird, not, not very common, like, you know, feeling lonely, feeling depressed, liking
Swagata:
those kind of posts, another indicator. So those were like good insights we got. Um, there were a couple of anomalies also people in a relationship and using less social media, but also Uh, we found out we’re depressed, but then it goes to another, another set of, uh, features like people who have said that, Hey, they’ve had like some, some childhood trauma or things like that.
Swagata:
So on top of the data that we had from their smartphone, we also did a survey where we had like fill them up some questionnaire. So there were some anomalies we got, like even though they have a childhood trauma. Very early, even now they feel they have those feelings, even though life is really good for them right now with the, they are in a relationship,
Swagata:
they have good academics, uh, they have good friend circles, still maybe sometimes demons of the past haunts them. So it, it was good, like we could get some anomalies as well as, you know, kind of generalize and get some really interesting insights from machine learning.
Hanh:
Mm hmm. Valuable. It’s so important to, uh, hone into the, uh, the mental health of teens, mental health of older adults. It’s much needed.
Swagata:
Yes.
Hanh:
And I think we’re, I’m seeing the delayed effects of COVID. Many of us, I don’t think have recovered from the loneliness and isolation from COVID. And uh, it’s, it’s very unfortunate.
Swagata:
I agree. I agree. We are seeing the side effects of COVID even now. Because, uh, people are still not getting used to the, the work from home setup or, or living in an isolated environment, they’re still struggling with it. So yeah, it’s interesting to, to have these kinds of tools.
Swagata:
So we, we understand them and help them out. Um, however we can.
Hanh:
So now, the methods that you describe, these machine learning methods, can it be used in other people or in other aspects, you know, other areas?
Swagata:
Absolutely. Yeah. I mean, you mean apart from mental health?
Hanh:
Mental health and teens? I, I believe it can, but I, I just thought, what do you think, like, where can we use that?
Swagata:
I mean, we can use, uh, the framework that we had for solving any sort of classification problem. Like, let’s say we want to classify whether a person. Uh, has diabetes or not, all you need to do is like change, change the data and, and little bit of tweaks here and there. But, but we can apply it to any area as such, wherever we are
Swagata:
trying to predict something.
Hanh:
Mm hmm. So, now, in a field as, um, dynamic as data science, continuous learning is huge. So how, do you approach staying ahead in this rapidly changing field?
Swagata:
It is indeed like in the past, you know, in weeks, certain new tools of fine tuning a large language model comes up. It is very difficult to learn everything. It is impossible. So what I try to do is I try to subscribe to some newsletters, which Kind of give me some concise information. I don’t need to learn every fine
Swagata:
tuning technique that is available. I just need to learn the top one or two. So I try to stay ahead by spending time in reading like newsletters or watching some videos, um, which is like 15, 20 minutes a day. And only when I see like, Oh, this is pretty valuable. I need to dive deep into it. Then I would, I would spend
Swagata:
some time and go deep, read some research papers, but I don’t. Think anybody has the time to, to learn every single thing that is coming because it’s just way too overwhelming. So if you want to do something quick and stay ahead of the, of the curve, just couple of ways is like network with people. If you don’t have time for
Swagata:
that, read a newsletter. There are lots of newsletters that give a concise summarized this week’s updates on AI. And if you find something which is interesting and caters to your needs or interest. Just, just read about it, and another way of learning a lot is attending conferences at events.
Swagata:
That is something being valuable for me because I attended a couple of conferences and what I learned from those conferences is currently what I’m applying at work. So that’s also, and I did, I couldn’t find that on online. So it’s a mix and it’s a combination of multiple things.
Hanh:
So, perhaps online courses. Corsica, is it Corsi? Did I say that right?
Swagata:
Corsera. Corsera.
Hanh:
Corsera. Corsera. Okay. There’s, well, when you say Corsera, I mean, that’s all you need to say. There’s nothing else in my mind. It’s really good, you know?
Swagata:
Yes.
Hanh:
Right. I love his work. Andrew Ng is just wonderful. You know.
Swagata:
He is amazing.
Hanh:
Yeah. So online courses like that and, um, networking events.
Swagata:
Absolutely.
Hanh:
Having one on one conversations like this to help kind of complement, supplement your knowledge and maybe even open some views that you thought one way at a point in time and it could make you more well versed, right? And I think the key is adaptable.
Swagata:
Absolutely. You said it right. Like being very adaptable. Um, not being, not living in denial like some people are not willing to accept that, hey, AI is important. So be willing to, to adopt it and adapt with it and learn from it. So yeah, adaptable is the right thing to do right now.
Hanh:
Mm hmm. And then echoing what you said earlier, it’s lifelong learning. And adaptability, those two are huge and you don’t even need to be a programmer. You don’t even need to know Python.
Swagata:
No.
Hanh:
You just need to know how to speak English.
Swagata:
Exactly.
Hanh:
Because that’s your prompt.
Swagata:
Yeah, that’s what AI has done. It has completely democratized tech for everybody. So all you need to do is just know English and, and, and you can, you can build AI products.
Hanh:
Mm hmm. Mm hmm. So true. So now, how do you balance the need for specialized knowledge and broad understanding of emergent trends?
Swagata:
Um, so I think one of the things is like I engage myself in a lot of initiatives apart from work because, you know, when you grow up in growing a career, a lot of times you’re doing a lot of managerial and leadership roles. So sometimes it’s difficult to get time to do like hands on and learn more things. In your field. So I try to get involved in
Swagata:
initiatives outside of work. For example, like the women in data chapter and talking at conferences about explainable AI. Those things kind of help me dive deeper and brushing up my data science concepts, because I’m not going to just talk about what is this concept, I’m going to show in practice how it’s done. So probably where you work on creating
Swagata:
a demo for people to understand, you need to understand it pretty well. So that’s how I try to, I try to balance it out, like at work, what you’re doing, and then outside of work, you can, you can participate in conferences and networking at the same time, taking. volunteering jobs, uh, so that you, you are, if people are looking up to you, then you have to be, you better be good.
Swagata:
So that’s how you.
Hanh:
Yeah. I think social media has opened up plethora of opportunities, right? Yeah. Conferences, virtual or in the physical location. And then it opens up for opportunities to join groups, webinars like this, or just listening in. So the opportunity to learn,
Hanh:
it’s out there, abundant. Now, of course, you got to choose what is trustworthy and um, what, uh, like, how, applicable in your niche, right? So, um, I just think it’s a great time to innovate, to learn, adapt, and just keep an open mind and be a lifelong learner.
Swagata:
Absolutely. It was a great time to be, be living.
Hanh:
It is. It is. It’s almost like Christmas morning, so to speak. You know, I feel that way every time I use. It’s like AI to create an image. My eye just like, wow.
Swagata:
Yeah. I cannot imagine not using these tools on a daily basis now. And like it was not there a year ago. So it’s amazing.
Hanh:
Yeah, it’s, it’s amazing, but a little bit scary because As you know, you know, the API, the open API was not working for, I would say half a day here in, in Eastern time.
Swagata:
Yeah.
Hanh:
Boy.
Swagata:
How am I going to work now?
Hanh:
Exactly. So I was gonna say like, what did you do to adapt for those four hours? Yeah. Yeah. Yeah. I’m with you. Okay. So as we kind of wrap up towards the end, do you have anything
Hanh:
else that you would like to add?
Swagata:
I mean, I was, I wanted to ask, uh, ask you actually, like, how, um, are you planning to, you know, spread awareness about, it’s a very interesting initiative that you have. If you don’t mind talking a little bit more about, um, what’s the scope of this and what’s the, how can we, I mean, me or other people in the community help you out with this initiative?
Hanh:
Yeah. Well, I’m very passionate of enhancing lives of older adults and providing tools, tech care for the aging population. And for the longest time, I’ve had the idea of providing an ecosystem where it’s not just, let’s say, senior living, senior care, because I want to focus more than just care. And I don’t want to group older
Hanh:
adults in one demographic because there’s a wide gamma of folks. I’m an older adult. I’m approaching 60. My needs are very different, 60s to 70s and 80s. But I want to. encompass all of the top issues that we will go through. For instance, affordable
Hanh:
housing, healthcare, senior living, legal issues, right? Saving for long term care, relationship with your spouse or dealing with loss, dealing with grandchildren. It’s a big scope that I want to address. But I think it’s more possible to keep it under one umbrella now that we have AI. I mean, if you think about it, open AI has this large language
Hanh:
model across the world, right? Older adults is one demographic in that. And I think it’s very possible to create one so that a senior will have their own AI assistant that will encompass all of those key components. And I’ve just mentioned a small glimpse. There are many. That I have not mentioned, but I think you get the gist.
Hanh:
I do. And I think the reason why it’s important, because we need to know that they, well, they need to know that, you know, they matter and that there is something, exactly. And that we recognize their concerns. Now when I say we understand, it’s hard to say that because, you know, we’re not 80s or 90s.
Hanh:
We live with our parents or grandparents, so we see their concerns and we want to ensure that they’re not alone. Mm hmm. There are resources out there that can help. And you know, you do feel alone when you go through this because you’re surfing on the web.
Swagata:
Mm hmm.
Hanh:
And who do you trust? Where do you go? That’s just one component, let’s say, affordable housing. Well, what about my health care? What about insurance or legal issues, you know, things of that nature. So my goal is to encompass that into one ecosystem.
Swagata:
That is beautiful. I love that. Let me know if, is there any way I can support you? I would love to do that.
Hanh:
Great. I do. I love to because we’re strategizing and um, you know, the key is enhancing lives. And I think it’s so important, like what we discussed, AI empathy, AI mental health, it’s about improving the quality of lives for people, you know, it’s not about building models and it’s not about just tech for the
Hanh:
sake of tech, it’s enhancing lives.
Swagata:
Absolutely.
Hanh:
And it’s not about. replacing people. It’s about freeing you up so that you do
Swagata:
things that you like.
Hanh:
The core, the heart, the soul of your business, right? Because I don’t like administrative stuff. I don’t like setting up meetings, you know, things of that nature. So that is my passion. And, um, with folks like you adding to the conversation. So I’m excited to continue our conversation after this.
Swagata:
Absolutely. I had a lovely time. Thank you so much for inviting me and, and I would love to support your initiative in any way possible.
Hanh:
Thank you. Thank you so much. All right. Take care. Have a great day.
Swagata:
You too. Thank you.
Hanh:
Bye bye.
