 Hi, everybody. My name is Shweta Ranganath. I'm a Senior Product Manager at Amazon Alexa, and I'm here to talk about product management in EI. What is product management in EI? Let's get down to basics of product management. Traditionally, product management typically means you understand a customer's problem and you define the business impact associated with the problem that could help your organization deliver value to the customer, but also making sure your organization is profitable. Then you come up with a product requirements plan, user stories, work with your software development teams and user experience teams internally to build the best infrastructure and the user experience to deliver the product to the customer. The major difference between a traditional product management framework versus a framework in product management EI is you will be working with a lot of data scientists and research scientists to build models and integrate the models into your product to deliver whatever value that you're trying to bring to the customer. Some of the fundamental differences is the way you can do this is, but actually the functions and team structure, which I just talked about at a high level. In a traditional product management, you would be the product manager, but you're the CEO of the product, and then you would have a software development team, you would have a user experience team, you would have some, if you're lucky you have BIEs and data engineers working with you to build analytics and tools to build feedback systems. Then the difference in product management AI is, you would have a team of data scientists or research scientists that are working on building models to make predictions or whatever the outcome that you want to bring out. Now, a lot of times when you think about product management in AI, it could be either internal or external products. A lot of external products usually are driven by a PNL. You're always bound by bringing business to the customer in some sense. Whereas product management in AI, you could be building products internally with AI, and deliver a lot of value to your organization by building optimizations, by building search engines, by building prediction mechanisms. I'll get to that in the next slide where I talk about some of the examples of internal products versus external products, and circling back to the point that I previously made, for products in AI, you may or may not have a PNL. If you're building an internal product, a supply chain product, you're actually still working in benefit of the organization by reducing the bottom line, or maybe reducing cost for the company. But you may not be actually bringing new business to the company. Whereas if you're probably building a tool in a social network where you're trying to increase users by using data signs and machine learning, then you're actually bringing business to the customers. This way, a product may or may not have a PNL when AI is included. Some of the examples that we can talk about in everyday life of products in AI that you may have already used, but you don't realize that there is so much of data and data science, and recent scientists working behind the scenes on this in addition to the product managers. One of the things that comes to my mind immediately since I work at Alexa is smart assistance. The technologies that you would use in all these different AI products would be slightly different, but they may be overlapping. When you think about smart assistance, you need to understand speech recognition, and then you need to understand natural language understanding. When you think about self-driving cars these days, you have a lot of data from the sensors that that's fed your time into some sort of a model, and then predicts maybe collision mechanisms or gives you alarms based on that. This is another example of how AI is used. You have smart vacuum cleaners these days. There are vacuum cleaners that help you train based on your existing data. You have again a lot of sensors built into the vacuum cleaners where you let it run and learn from the data that it has, and then it uses that data to train itself on how to clean going further. Social media feeds is another example. So you go and search something on social media today, and you have a ton of recommendations and suggestions in your feed based on your historical searches. So the way they use this, the way it's done is you kind of use the existing data based on your search and then you build models to kind of predict what you would actually like to see in future. And pretty much any e-commerce platform today uses recommendation systems. So they look at your previous history, see what you like, and then make your life easier by kind of recommending on what products you would like. And I think personalization is very similar to that. It could be things like your music choices or your video choices that you would like to see based on what you've previously seen and use that data. Now coming to the internal tools that I actually talked about is automations in supply chain or forecasting and pricing. These are some of the internal AI products that you could build to actually improve the productivity of your company or your organization. Now, what are some of the key things that's important for actually fundamental things that's really important while designing products in AI? The three things that come to my mind immediately is data and data acquisition for training models. So when you're building a model, it's really, really essential to make sure you have the right seed data to kind of cover enough of the, almost to say to replicate the users that you're building for. And another thing that's key for AI is kind of getting the data itself. So data acquisition itself is a huge challenge by itself in AI in general. So when you kind of go into this building AI products, I would say that the key thing for you is getting the data for training your models. And the second thing is the means of getting the data, I would say. The second thing that I can think of immediately is privacy and security. Now, a lot of these products that you're building is based on customers' private data. So you wanna make sure that while you're designing your systems, your AI products and also when you're acquiring the data, you are giving the utmost importance to prevent any kind of data leaks or data expiltrations. So you wanna make sure you're protecting the customer's data by choice. And you do that by designing your products all the way from building the right infrastructure and also getting the data from the customers. You're also building the products and making sure your experience is not actually revealing any kind of private data to the customers. So you wanna make sure you're end-to-end, you're created to deliver the product is completely privacy-proof and security-proof. The third most important thing, in fact, I would say this is probably the most important thing while building an AI system or an AI product is the bias and inclusivity. So because so much of AI happens behind the scenes, you wanna make sure that your data, your AI, your model itself is not biased in any way. And the best way to do this is to make sure your data, your seed data, which goes back to the first point, is actually kind of covering all the different kind of variations that you would see in a real-time data that you would build a product for. And of course, inclusivity. So you wanna make sure you're in addition to making sure data and your models don't have bias, you wanna make sure the way you've designed your feature does not actually exclude any kind of user group from actually by using the product. So you wanna make sure you're building all of these bias, you're making sure you're very conscious about your choice, making sure there's no bias and making sure there is a lot of inclusivity, both when acquiring the data for building the models and also while designing the features itself. Now, one thing that's commonly asked, I would say, is how do you kind of get there? How do you become a product manager in AI? So I can talk a little bit about my own journey in terms of how I got here, and then I can talk about ways that will actually help you get there. So I have a background in engineering, so it actually helped me a lot to kind of already have that back on. And once I finished my engineering, I did my master's in electronics and communication where I did a course in artificial intelligence that kind of gave me a foundation. But although that helped me, it didn't really understand, helped me understand how do I implement AI in a real world, right? So it took me a couple of years to do an engineering role where I worked a lot in statistics with engineers and then helped me understand how to use data to do predictions and mechanisms. And once I finished that, I think I was so involved in my engineering role and I had to kind of pull back and understand where engineering connects to business and how you could use the data that you have to actually deliver business value and understand customer problems. So it's really kind of having that bridge of understanding where you deliver business value, who your users are, what problems you're solving for the users. And you're not over-engineering a product to actually deliver value to the users. So you want to make sure that you're using sophisticated mechanisms and systems to actually build products, to using AI to build products, but you shouldn't be over-engineering or use so much data that it's not required. You could deliver value to the customer by doing something simple, by actually delivering, we're building a model with a regression, like a regression model, which is really light-paid, does not require a lot of training data. I mean, that does not require a lot of, you know, GPU time to actually deliver your data, delivered results in real time. Whereas you could build really sophisticated systems that takes data from real-time processes and then puts, gives you output back to kind of deliver to the customer. Now, just kind of going back to my story on how I got there, I did a course which helped me kind of build more knowledge in marketing and finance, because I think it's really essential when you're trying to build a product, whether it be AI or not AI, it's really essential to understand how you're, you know, how do you estimate the value that you bring to the organization as well as the value that you bring to the customer. So I would definitely say, if you are actually thinking about becoming a product manager, it's really important to understand how a P&L works. The other thing that helped me a lot is marketing itself. So I feel like a lot of vision-setting in a product because a product that when you think about a product, it's not just, hey, you build a feature and you move on. You're thinking about all the way from your customer's point of view on how you can set a vision and kind of develop a multi-year roadmap which has milestones. So you think about how you build a basic product, your MVP, and then you kind of add feature enhancements and experiences to kind of improve the experience for the customer. So you have to think about how you do your vision-setting. So I feel like a lot of marketing knowledge helps you just build the vision in that sense. And it also helps you understand how do you go to market with a product? So if you're building a product that has a P&L and you're expected to deliver, you're expected to have incremental revenue on those kind of goals and your product, then I think it's really essential you understand how you go to market. So your pricing, your placement of the products, your go-to-market plans, all of these things are really essential. And for me itself, I think taking courses in marketing helped me a lot. And of course, I did a course in data science which helped me even more. I think one of the key things is how do you understand what is required to actually build a model? And I think people assume that it's easy to understand, but there's so much of basic understanding. So you need your saved data to actually train your model. You need to make sure you're building, you're actually picking the right kind of training method. So you have different kind of training methods. Like you use decision trees, you can use regression mechanisms. You can use scan and models. So the different type of models that actually help you deliver different kind of results. So if you are building a recommendation in an engine, it's a different kind of model you would choose. Or if you're trying to build a personalization engine, it's a different kind of model that you would use. If you're just trying to predict house prices in your neighborhood, again, it's a different kind of model you would use. And also one major thing that I've learned over the years working in the AI world is a lot of companies that are at the forefront of AI. You use ensembles, which means that you use a bunch of different techniques to build a model. And then you basically merge them into one single model. And then you use that model to deliver your results. So you have like a lot of sophistication there, but it also means a lot of GPU time when you run models that are heavy. So you have to think about the cost that goes into building a model while you're building the product that delivers a value to the customer. And of course, I think one thing that I did was I actually did a stretch project early on in my career when I was in engineering role and product management, where I actually defined a product policy for customer engineering solutions. So having that extra actual real work experience for moving on to a role full time really helps you in addition to actually upscaling yourself with academia learning. And having a mentor helps a lot. So for me, I had a bunch of different mentors and I feel like the mentor that you choose is really important. Cause I think there are mentors who help you kind of navigate your career, but you need help, you need mentors who actually give you effective feedback. So picking a mentor who you work with regularly, who is maybe a senior role in a product management forte, will really help you kind of getting the right feedback on building the right skills that you're missing right now to get to a product management role. And I feel like when you're done with all of this, I mean, you're kind of ready to move on to a product management role in AI. You should definitely do a lot of mock interviews to understand where you stand and what's missing in your current profile. So a lot of times you've actually done work that can be placed as product work, but I feel like it's really understand what product management means and how you've fit into giving the right kind of examples to give the interviews, to make sure you actually put the nail there and get the job. So this kind of summarizes, I think the talk about product management in AI. I'm happy to take questions outside, obviously, of this time here. And my bio should be linked on LinkedIn and I'm happy to walk you through any kind of questions that you have or disconnect for a social coffee chat if you'd like. So I hope this was helpful for budding product management managers who are looking to get into the AI technology in general. And just helping learn the basics of product management in AI. Of course, I think the deeper you go and think more about how you build product management products in AI, there's a lot that you can talk about all the way starting from the kind of services that you choose to build AI products and how you integrate that, how do you kind of get the data to build your seed model, how you actually work with scientists, research scientists and data scientists to deliver value. So I feel like one of the things that's more essential is while you're actually writing a business, you're defining the business impact and thinking about how you have a vision for a product, it's really important to work very closely with data scientists and research scientists because I think feasibility becomes very important in executing a product in AI. So I feel like all of these things are so tight knit. And if anybody is really interested in learning about this more in detail, I'm super happy to connect offline. You have my contact details just below here. You'd also have my profile linked to this webinar where you can reach out to me on LinkedIn or a message if you'd like. And then again, I really enjoyed talking about product management in AI and product management in general. I really hope this is helpful to people who are looking to get into a career in product management in AI. And thank you very much for staying on and listening in on this and have a good rest of the day. Thank you very much, bye-bye.