This episode features an interview with Vaibhav Kulkarni, Head of Data Products & Infrastructure Engineering at PepsiCo. Vaibhav has spent over ten years working his way up from an engineer, to a founder, and now to a thought leader in the tech industry. In this episode, Vaibhav talks about the transfer of e-commerce to the cloud, what data transformation looks like on a massive scale, how to increase your ROI, and much more.
This episode features an interview with Vaibhav Kulkarni, Head of Data Products & Infrastructure Engineering at PepsiCo. Vaibhav has spent over ten years working his way up from an engineer, to a founder, and now to a thought leader in the tech industry.
In this episode, Vaibhav talks about the transfer of e-commerce to the cloud, what data transformation looks like on a massive scale, how to increase your ROI, and much more.
--------
How you approach data will define what’s possible for your organization. Data engineers, data scientists, application developers, and a host of other data professionals who depend on the Snowflake Data Cloud continue to thrive thanks to a decade of technology breakthroughs. But that journey is only the beginning.
Attend Snowflake Summit 2023 in Las Vegas June 26-29 to learn how to access, build, and monetize data, tools, models, and applications in ways that were previously unimaginable. Enable seamless alignment and collaboration across these crucial functions in the Data Cloud to transform nearly every aspect of your organization.
Learn more and register at www.snowflake.com/summit
RODC035
Steve Hamm: To be vibe. [00:00:00] Welcome to the podcast. It's great to talk to you again.
Vaibha Kulkarni: Thanks, Steve. It's my pleasure to be here.
Steve Hamm: that's great. Now everybody knows PepsiCo, at least the products, the brands, but it would be good if you could start off by describing the business and all of its dimensions, and then focusing on the e-commerce unit where you work.
Vaibha Kulkarni: Certainly. PepsiCo is one of the largest CPG companies in the world. Uh, it was founded by merger of Pepsi Cola and Frito-Lay companies. So we have a very diverse portfolio of food and beverages products. And to name a few , uh, you know, Pepsi Gatorade lays Quaker, Tropicana Doritos. We have close to 23 brands , uh, with billion dollar valuation each.
We are continually expanding our portfolio to go beyond the bottle , uh, and find ways to deliver beverages free [00:01:00] from single use plastic. Uh, and our soda stream business is definitely enabling this initiative. Oh , uh, a fun fact. Uh, our products are enjoyed by consumers more than 1 billion times a day in more than 200 countries and territories around the world.
Now , uh, speaking about e-commerce , uh, e-commerce ARG is relatively new within Pepsi-Co. We started like five, six years ago. Uh, we are divided into several teams that are focused on specific areas like supply chain. Uh, we have D two D uh, DTC, B2B initiatives. Uh, we have marketing sales strategy and technology groups within e-commerce as well.
And we are also building several global products , uh, like ROI engine that goes beyond e-commerce.
Steve Hamm: Great. We'll talk about that a bit later. Hey, so what is your personal role in PepsiCo in PepsiCo's overall technology and data strategy and in particular, what's your role in [00:02:00] e-commerce?
Vaibha Kulkarni: At PepsiCo e-commerce, I'm heading up the data products , uh, engineering and infrastructure engineering groups. Uh, we are responsible to , uh, build Kelly bowl and extensible data platform, project train the cloud. Uh, my, my, my group is also working on multiple global data products, initiatives , uh, focusing on data and application engineering and then analytics capabilities.
Uh, we work out with many, many, many of our stakeholders , uh, here in the United States and globally. Uh, we, we already launched several products, including ROI in gen, in North America sector last year. Now, now we are expanding our capabilities to launch it in our global sectors like Europe, APAC, and other regions. On the other hand, my infrastructure engineering group is focused on , uh, building cloud infrastructure , uh, tools and systems that that's empowering our hundreds of technologists within PepsiCo e-commerce.
Steve Hamm: Um, now you have led technology teams at a number of companies, including a startup [00:03:00] called miles. Please talk about the most important management and leadership lessons that you've learned there that you're put into use at PepsiCo.
Vaibha Kulkarni: I was working at miles , uh, startup , uh, before joining PepsiCo , uh, miles is a multi-modal transportation reward mileage program. Uh, basically as a user, all you need to do just install the mobile app and it would detect all of your trips automatically. Without human intervention. Uh, so be it car trip , uh, train bursts, or, you know, activities like walk, run bike , uh, you basically get the miles for these trips.
Now you can redeem your miles for various rewards. Uh, as a founding member of files, I was leading engineering and data science teams there. You know, as a tech leader in a startup, you constantly have to think about how to deliver value to the business faster. Uh, you do not want to over engineer things from day one, especially if you are literally [00:04:00] bootstrapping discharged up, you have to focus on delivering the features at the same time.
You also have to use the resources optimally, considering the limited funding. I was part of the journey at miles, where we grew from four to five people to 25 plus within a year and half as leader celebrating small wins is very important. It could be a spot reward or recognition. This will keep your team motivated.
I also firmly believe that you have to give spirit place and autonomy to your team members to get the best out of them. Overall as the leader in Pepsi-Co e-commerce my focus is to create a very empathetic, inclusive and collaborative team culture by empowering team members.
Steve Hamm: basically, you're taking the lesson to learn, done a small company and you're applying them in a [00:05:00] giant company, but for a team, I mean, it's a much bigger team, but it's , they, they translate well,
Vaibha Kulkarni: yes, definitely. And then one of the aspect is , uh, when I joined Pepsi-Co e-commerce I was told e-commerce is kind of like a startup within the big arc. And , uh, I definitely agree with this.
Steve Hamm: it makes a lot of sense. Now within PepsiCo's e-commerce unit, I know you have a number of multi-disciplinary teams, combine engineering, data, science, business, side, people, analysts marketing, and product management people. So it it's a really diverse set of people with different goals. Though they share a common goal.
So how do you set up and operate these teams so that the different domains are kind of synchronized and you get good results out of it?
Vaibha Kulkarni: We work in very highly integrated cross-functional team environment. Uh, we have , uh, we, some, we call it pod structure , uh, for different projects. [00:06:00] And , uh, these parts comprises , uh, basically they comprise of , uh, engineering team, data science team , uh, product, business, marketing stakeholders. Now these parts have daily stand-ups weekly or biweekly , uh, you know, replenishment meetings to ensure we are prioritizing.
Thanks to deliver value to the business. Of course, this part structure is evolving as we speak since we are also working with , uh, know multiple global sector teams. Uh, now we not only have to deal with one stakeholders, but imagine , uh, you know, you have, we have the Europe stakeholders, we have the stakeholders from APAC and within each sector we have stakeholders from Grisha.
We have stakeholders from , uh, particularly if we're talking about Europe , Uh, you know, UK pain, et cetera. So it's really important , uh, that all teams are in the sync , uh, and , uh, to deliver projects successfully. Uh, sometimes over-communication is good when you work with that many teams together. Okay.
Steve Hamm: Yeah, that makes a lot of sense. Now PepsiCo [00:07:00] is very aggressively launching new applications, both internally and externally. And I know that recently you launched pantry shop.com. This is a platform for consumers, and it came out last year. If you could describe that initiative and the role in it for data and data analytics, that would be very helpful.
Vaibha Kulkarni: Definitely as the world adapts to new ways of living and working , uh, being consumer centric is more important than ever. Uh, consumers are increasingly turning online for their food and beverages needs. Uh, we launched this initiative, Patrick shop.com our direct to consumer initiative. Where a shoppers can quickly and easily order their favorite products and get it delivered to their doorstep.
We devloped the site. Within less than 30 days in response to consumer evolving needs. During these uncertain times last year from data perspective , uh, [00:08:00] we do some user level analytics to better serve our consumers. Uh, at the same time, we mainly use this data for forecasting and supply chain planning purposes.
Steve Hamm: Um, now you mentioned at the top that one of your major projects today is your ROI engine. And I understand this is an in-house data analytics engine for evaluating marketing and advertising campaigns, and also for measure I, and for measuring how effective they are. So I want to know when and why did you launch this initiative?
And also, I mean, this addresses all sales channels and not just e-commerce right.
Vaibha Kulkarni: Uh, with consumer and customer landscape evolving. Uh, we launched our engineering sheriff a couple of years ago. Uh, this was our efforts of building an in-house media optimization tool or media measurement tool. Uh, the ROI engine allows us to make smarter decisions in measuring the campaign effectiveness and return on investment. Uh, it's a multi-variant machine learning, modeling based tool and yes. You are right. The ROI engine [00:09:00] addresses, not only just e-commerce, but we also address many other channels such as convenience, grocery drugs, retailers, et cetera. As part of this project, as you know, a big arc like Pepsi CO's spends a good amount of money towards marketing.
And this tool helps our marketers answer some of the questions that Chaz. You know, quite recently, what's the effect of COVID on sales. Uh, how did, how did , uh, let's say my TV campaign perform or , uh, you know, uh, it also gives recommendation to our market is like, is it better to put next dollar in digital?
Uh, our TV. So for example, let's say you're spending like , uh, just coming up with the numbers, you know, 20% on , uh, off your marketing spend on TV , uh, you probably would just want to spend 15% on TV and. Uh, maybe preventing 5% on other digital media when these are these , uh, questions , uh, that , uh, this , uh, tool helps answer.
Right? Of course it is. Now it can affect ROI as well. Uh, you know, their holiday or , uh, uh, you know, uh, uh, And then , uh, we decided to build this tool in house because , uh, previously [00:10:00] we had a system that ran less frequently. Uh, we had to wait several months to get the results delivered. Uh, and many of the things were literally black box.
Now we have built faster.
Steve Hamm: by black box. You mean you, you, you got a result, but you didn't know how you got it. Is that what you mean by that?
Vaibha Kulkarni: that is correct. So, uh, it was outsourced to our vendors and then all we saw just the originals. Right? So, uh, now , uh, uh, we have built this tool, which is faster, more granular, more frequent , uh, and then , uh, with the better coverage , uh, we are delivering results to our marketers faster. And they can quickly act upon it.
Uh, dish finally , uh, we have complete ownership of data set and we can also run more analysis on top of it to make the better decisions.
Steve Hamm: Yeah. The idea that you had to wait months to get the results and understand how successful your campaigns were is just incredible. That just sounds kind of like, Oh, so 20th century, you know, but I think one of the things that has [00:11:00] enabled this is the movement of data to the cloud, right?
Vaibha Kulkarni: yes, certainly, definitely. That helped a lot.
Steve Hamm: Yeah. So when and why did PepsiCo start migrating its data to the cloud?
Vaibha Kulkarni: sure. So, uh, the cloud migration initiative , uh, I would say, was started a few years back , uh, as a company, we collect tons of data across our different properties and products. Uh, so moving from on-prem infrastructure to cloud was of course no brainer as it offers a lot of benefits , uh, you know, such as, you know, security and a reduced risk of data loss , uh, than the infrastructure we can easily , uh, able to deploy.
We are able to deploy. Now we're containers to Kubernetes clusters in cloud, which makes it easier to scale our services and applications. Of course, the computation on demand , uh, that gives us a capability to do , uh, faster analysis , uh, you know, to , uh, basically , uh, uh, D demand computation to train our machine learning [00:12:00] models effectively.
Oh, with the cloud adoption, we are definitely becoming faster, stronger, and better to achieve our PepsiCo's broader vision of be the global leader in kind of Indian food and beverages by winning with the purpose
Steve Hamm: I'm sorry, by winning with what?
Vaibha Kulkarni: purpose.
Steve Hamm: Winning with purpose. That's an interesting idea. So that's like the mission of the company.
Vaibha Kulkarni: Yes, that's our broader version of the company.
Steve Hamm: Yeah. Oh, I like that. Great. Now, whenever companies do these major transformations and these major migrations of data to the cloud, it's a big deal because it's changing kind of the, it model, the data model and some real business processes.
What challenges have you encountered in the, in this migration or through this migration and how have you overcome them?
Vaibha Kulkarni: Fortunately, within PepsiCo e-commerce, we didn't face a lot of challenges migrating [00:13:00] to the cloud. As we were relatively a new organization, our cloud adoption was rather smoother, but of course there were some other challenges. I mean, I've seen this not only within PepsiCo, but also some other big carps I worked at, you know, a lot of data is sitting in Excel and PowerPoint deck.
Uh, we of course , uh, now the importance of , uh, of , uh, the data get, we get that data onto the cloud to deliver actionable insights and we all are working towards that effort. Okay.
Steve Hamm: Okay, sounds great. So it's. It's not so much challenging as there's, there's a lot to do because you've, you've got to, you've got to get it out of those applications and those databases that's already in and make it available to the cloud in a new way. It sounds like. So that sounds like a real data engineering challenge rather than anything else.
Vaibha Kulkarni: Yes, definitely. And then a lot of the data sources , uh, you know, we have to collect, we first have to discover these different data sources, where are [00:14:00] they? Are they, they could be , uh, you know, uh, uh, the company as big as PepsiCo with 300,000 employees, you know, it's, it's really hard to, it could be really hard to , uh, find the data's forces, the internal PepsiCo data sources as well as external.
Data sources. Uh, the external data sources has become , uh, the exploration of external data sources has become a little easier nowadays with, you know, uh, technologies such as snowflake marketplace, et cetera. But yeah.
Steve Hamm: Yeah, that's great. Now I wanted to explore a little deeper on this, on the data cloud. When and why did you begin working with snowflake?
Vaibha Kulkarni: We've been using snowflake , uh, for more than two years now, at times we were looking at multiple cloud data warehouse solutions , uh, and snowflake particularly stood out. I think snowflake offers very unique data cloud solution with their , uh, architecture. You know, they've separated. The compute and storage layer , uh, has several benefits apart from performance and speed.
Uh, you know, you're, [00:15:00] you're ready for a multi-cloud environment really , uh, within PepsiCo. Uh, we have teams using multiple clouds , uh, for different purposes, and it was important to make our data available across clouds, to work on insights and snowflake enabled there's through their, one of their features , uh, where you can literally basically set up multi-cloud replication.
Uh, the replication across cloud with a couple of clicks. Uh, we also did a benchmarking at times, and then , uh, uh, it was best choice for us within PepsiCo. E-commerce considering the cost and performance snowflake we were offering.
Steve Hamm: well, you, so you've had stuff like for a couple of years, and I imagine the way you use the technology has evolved. What role does it play in the ROI engine? And also does it enable improved data sharing? It seems like those are, those are two key questions for you.
Vaibha Kulkarni: Yeah , so, uh, snowflake , uh, Uh, first of all, allowing teams to build insights very easily. Uh we've so far we [00:16:00] have , uh, ponders of data pipelines that we built , uh, using our data platform and we use snowflake , uh, on our day to day basis , uh, to , uh, to basically store these data sets in the cloud. Uh, and snowflake is powering , uh, multiple applications within PepsiCo already.
Uh, we have , uh, Tableau the I analytics powered by snowflake. We also have some internal web applications powered by snowflake. Our data scientists are running their models on top of the data that we're storing in snowflake. Uh, The snowflake data sharing feature, particularly , uh, uh, I mean, I want to talk about it , uh, specifically regarding ROI engine, as you know, for the , uh, uh, as part of the ROI engine, we are collecting data from now 60 plus data sources.
And , uh, once we produce , uh, the actionable insights, we store them back into snowflake. Now we have to, we have to share these results across different organizations within PepsiCo. Uh, and , uh, snowflake is enabling this by , uh, you know, uh, this [00:17:00] feature called data. Sure. Uh, where we are able to quickly share a database , uh, our table with these teams, and then they don't need to worry about building these data and drink pipelines.
They always get the most latest and up to date data available to them at any given time. Uh, and then of course , uh, uh, we w we, there are multiple products that we are building on top of snowflake, not only just ROI engine.
Steve Hamm: okay. Well, ROI engine certainly sounds like a great product, a great capability, and it really illustrates one of the core things about the data cloud, because. You you're bringing together all these different groups, you know, you're bringing together your marketing people , your, your product, people, your data engineering, people, your analytics, people who previously, and then all the different lines of business previously, they probably had all their data separated, but only because of the cloud.
Can you, can you get it? [00:18:00] Can you integrate it and manage it and, and makes it actionable. It seems like that's really the heart and soul of ROI engine, correct?
Vaibha Kulkarni: Yes, definitely. So, uh, Having all of these data sets together in one place is very beneficial. And we already seen servings in our marketing spend last year because of this tool, specifically the ROI engine. And we are able to make choices at many different levels. Uh, you know, from AOP levels, strategic choices, where do we place our bets?
Uh, you know, uh, uh, we are able to make some tactical and strategic decisions by doing top level analysis. Uh, The ROI engine already allowed us to increase our digital penetration in certain brands by more than double digits. Right. And , uh, of course we are able to do the informed decision by knowing where our consumer mindset.
So, yeah.
Steve Hamm: Yeah. And I imagine in the midst of COVID when suddenly so much was uncertain in society and health and in the business world, having something like ROI engine. I [00:19:00] mean without that you would have had a very hard time responding to those changes and to kind of like the mysteries of where things were going and what was happening.
So it seems like that was, that was a very smart move, even though you didn't obviously anticipate COVID it came in handy. So that's really interesting. Yeah. And the, the , um, ROA agent, just so I'm clear on this, it's used within e-commerce, but also in other parts of PepsiCo, correct.
Vaibha Kulkarni: Yes. So, uh, ROI engine, particularly , uh, so like I said earlier, e-commerce is just , uh, one of the channels that we deliver results to, but we also deliver results globally across different channels to different sectors globally. So this is way beyond, uh e-commerce as well. We have multiple teams involved , uh, from, in PepsiCo global , uh, side of stuff.
You know, we have the , uh, the global insights team. We have global marketing and media teams involved in this project.
Steve Hamm: Yeah. Yeah. That makes total [00:20:00] sense. Now I understand that. You're I understand that you're beginning to use Snowflake's data marketplace for acquiring and sharing data. So how does that change the game for you?
Vaibha Kulkarni: So we recently started using a snowflake data marketplace. And let me tell you, it's very easy to get started with , uh, you've have the data available to you very quickly. There are tons of data sets available in the marketplace and many companies are publishing newer data sets very often. Uh, we are actually using several data sets through marketplace.
Uh, so for example, you were talking about COVID , uh, just a little , uh, uh, earlier , uh, and , uh, we are, we are, we were able to get the COVID data set , uh, well, uh, for, for, for the ROI engine project, specifically through snowflake marketplace very quickly. Uh, uh, at the same time, we also looking at some other data sets in a snowflake marketplace, like foot traffic data and , uh, you know, uh, the weather data, et cetera.
Steve Hamm: So. Looking ahead. What are the most important trends in cloud computing and in data [00:21:00] analytics that you see coming in the next year or so
Vaibha Kulkarni: More companies are realizing the real value of the data, especially all the things we could do with cloud computing power. We will see more and more need for data technologists. Uh, I think this, this is, this is the penetration that we all seen back in nineties or 2000 when all of the companies were , uh, running to get their businesses online.
Uh, right. Uh, and , uh, as the things are still evolving in data space, there are so many roles in in this field. For example, we have now data scientists, data analysts, data engineers, machine learning engineers. Then we have , uh, I also heard like full stack machine learning engineers. There's tons of opportunity in data space.
We already seeing some impressive startups and tech coming up in, in this pace , uh, to name a few, you know, Monte-Carlo data. Uh, there they are in data quality on observation, observability space. Now a DBT, an ETL tool cube flow in MLR space. I think, I think we just get [00:22:00] also getting just started with the data, share across the companies through the data cloud and snowflake particularly is playing a key role here, but their data share , uh, capabilities.
I feel a lot of companies will take advantage of such features cutting down their data, transfer costs and development time. Uh, I think companies will be able to sure. And ingest the data without needing to build this big ETL processes through , uh, the snowflake data share capabilities specifically.
Sorry, can you hear me? I think you're on mute. Yeah. Now you're good.
Steve Hamm: Okay. So they'll, there'll be less need for ETL because the data has already been loaded in prepared. Is that the idea? Yeah.
Vaibha Kulkarni: That is correct. So, uh, again, with the data share feature, you don't really have to worry about building these , uh, ETL pipelines. Uh, when company companies shares their data sets, we can directly consume it without , uh, you know, building these data pipelines.
Steve Hamm: so you kind of skip it, skip a big complex, you skip a big, [00:23:00] complicated step. That sounds really smart. So now I'm going to ask you to put on your visionary cap for a minute and look out five years or more. How do you see data? How do you see the data cloud impacting business and society?
Vaibha Kulkarni: So data driven decision making. Definitely this will enable companies in different fields to make informed data-driven decisions. Companies will further find new ways to generate more revenue , uh, produce some actionable insights and find new growth opportunities. And I'm very optimistic about , uh, you know, future of the data and AI technologies, particularly , uh, we already seeing the benefits of AI knowingly or unknowingly , uh, right now , uh, I remember I used to work in Goshen technology around five years ago, which is by the way, leader in digital coupons and omni-channel digital marketing.
And one of the projects I was leading , uh, there was , uh, an applied machine [00:24:00] learning. We were trying to recognize and classify retail, product images using mobile camera. So basically , uh, it could identify if a product is diet Pepsi or Gatorade or laze just by pointing a camera at it about same time. Uh, there was a need there, there weren't some news about machine, or I would rather say.
The artificial neural networks like convolutional neural networks were just outperforming humans in identifying the images accurately. It's very fascinating considering all your computer sees. You know, just, it's just a bunch of numbers in a metrics. So let's say if it's a hundred by a hundred pixel grayscale image, all your computer sees is a hundred by a hundred pixel of metrics, but each number in the metrics representing zero to two 55 color range.
Anyway , uh, uh, we all are seeing emerging AI applications. In different industries, you know, self-driving cars, Amazon [00:25:00] ghost tours, identifying cancer. I think these things will become a norm in future. And I'm quite excited about the impact of the data and tech , uh, can have.
Steve Hamm: The, I heard recently that you've been doing some meetings in the virtual office. Now we're all used to using zoom for our meetings, but you are literally using a VR app.
Tell us about that. What is it and how does it work?
Vaibha Kulkarni: Oh, sure. Uh, so I have this FIAR headset , uh, and I use , uh, I use this app called immersed VR on it, which is pretty cool by the way , uh, you can work with your colleagues in a virtual. Office environment. So basically you are in the same Warshall room. You can literally see your colleagues setting besides you.
Of course, in their virtual avatar. You can also have multiple virtual monitors to work with. Uh, you know, it connects with. Connects to your laptop. And then of course there is a virtual webcam, which by the way, shows your, [00:26:00] your avatar and virtual office background in your zoom meetings. That's fun. And at the same time , uh, you know, you're getting, you're getting work done.
Sometimes this is what needed and , uh, for this very unfortunate pandemic situation.
Steve Hamm: So it has a little bit of spice to life and you can also get your work done. Like a double benefit. That's pretty cool. Yeah. Yeah. I , um, I love that example and I just think, you know, we're going to see a lot of new collaboration technologies and approaches. They're just going to be incredible over the, over the next coming years.
You know, the COVID thing kind of set it off to kind of set off a new wave of innovation there. As a matter of fact, I was listening to Eric Jaan , uh, Of of zoom talk earlier today. And he said he calls Silicon Valley. He doesn't call it Silicon Valley anymore. He calls it collaboration Valley because that's where he thinks a lot of the innovation is going to be going on in the coming year.
Yeah. So this has been a really cool conversation. I love [00:27:00] that. I also think your ROI engine is really powerful. You know, I can see how practically any business of any kind that does advertising or even, or even aggressive marketing. They need something like that. And I, and it seems to me that that's something we're going to be hearing a lot more about from a lot of companies.
So, thank you so much for your time that I think has been really illuminating.
Vaibha Kulkarni: Thank you, Steve. It was great. And my pleasure talking with you.