AWS Financial Services Symposium | Andrea Anderson | S&P Global & Ashley Videtto

Clips
News
More from AWS Financial Services Symposium

Andrea Anderson

Executive Director of Real Time Analytics and High-Performance Compute Platform Capabilities

S&P Global

Ashley Videtto

Solutions Architect

EPAM Systems

Cloud-based energy data analytics enhances efficiency in oil and gas with Energy Studio: Impact

Energy data analytics is rapidly revolutionizing various sectors by offering real-time insights and enhanced decision-making capabilities. A standout example in the oil and gas industry is Energy Studio: Impact, a cutting-edge platform that is transforming decision-making processes.By leveraging swift geospatial processing and advanced data visualization, Energy Studio: Impact embeds significant data to provide a comprehensive, data-enabled experience

play_circle_outline Need for quick decision-making in the upstream oil and gas industry

play_circle_outline Collaboration between S&P Global and EPAM for a cloud-based application

play_circle_outline Use of AWS services for data storage, analytics, and performance

Info

Andrea Anderson | S&P Global & Ashley Videtto | EPAM

Andrea Anderson

Executive Director of Real Time Analytics and High-Performance Compute Platform Capabilities S&P Global

Ashley Videtto

Solutions Architect EPAM Systems

>> The AWS Financial Services Symposium, presented by theCUBE.

>> Hey, everyone. Welcome to theCUBE. We are here in New York City. I'm John Furrier, your host of theCUBE, and we are here at AWS Financial Symposium in Manhattan. It's part of tech week. We got a lot of great coverage. We're going to be covering a lot of range of interviews. We've got a great segment here. Andrea Anderson, Executive Director of Real Time Analytics and High Performance Compute Platform Capabilities with the S&P Global Commodities Insights. Thanks for coming on.

Andrea Anderson

>> Absolutely.

>> Ashley Videtto, Solutions Architect at EPAM. You guys did a great solution. Thanks for coming on.

Ashley Videtto

>> Thank you.

>> Ashley, let's get into it. So you guys are working together. You guys had a need for a solution. You guys delivered it with AWS, built it out, stood it up. Pretty killer solution. Let's get into it. What are you guys doing with this product?

Andrea Anderson

>> So with Energy Studio Impact, before we actually even had the name, the need that we had really identified, and we were primarily at the time working with the upstream oil and gas industry, we recognized that they had a need to make decisions very quickly. Especially in that industry, in that space, the ability to identify a deal or an asset that you want to acquire very quickly is really important. And to do that and to make solid decisions, you need a ton of data. A lot of companies don't have the internal infrastructure to grab all the data necessary and then give their users the ability to quickly convert that data into insights and decisions. So we wanted an application that was cloud-based, that had all the breadth and depth of data that we have at S&P Global available at their fingertips through a modern interface that gave them the capabilities to collaborate, create, and customize.

>> The Energy Studio is what that's called, the product?

Andrea Anderson

>> Yes, exactly.

>> So what was it? Clean sheet of paper, you have all this data laying around and say, "Hey, how do we roll it quickly to make better decisions?"

Andrea Anderson

>> Exactly. How do we get into a place that is utilizing a BI tool type of workflow, right? I want to create charts, graphs, I want to do geospatial processing. I want to start moving into the world of kind of predictive, utilizing AI ML and getting them to a place where they can take kind of this basis of data and turn it into very quick click, click, click, and understand your

>> What was the demand driver on your side? Was it more of the deals? Was it more we have this data? What was the key driver?

Andrea Anderson

>> Yeah, the key driver, so we're ultimately a data provider at the end of the day. Users have access to all of this data, but the trouble is, how do we consume it and how do we actually convert it into insights very, very quickly because we're talking about hundreds of millions, if not billions, of rows of data. So we really needed that interface and that delivery mechanism that they could quickly and easily access it.

>> Ashley, you're solution architect filling in for your colleague. Thanks for coming on. I appreciate that.

Ashley Videtto

>> Thanks.

>> Shout out to him. Hope everything goes better there. Quick question for you. This is a classic use case of moving fast and with the cloud, obviously, the cloud's great, but with generative AI, you want get it right. You want to set the table, deliver the solution, but put it in position to take advantage of all the generative AI stuff coming around, which everyone's buzzing about. How did you guys approach this? EPAM's doing a lot of great work with your clients, but in this case, take us through how do you look at the problem? How did you guys tackle it?

Ashley Videtto

>> Absolutely. So Andrea and her team were really key at identifying the high-level vision and the key goals that they wanted to accomplish. They know their industry very well, so she had all the requirements of the application laid out for us, ready to deliver. What was unique and interesting about this particular use case is not only did they want to deliver it fast, they had very aggressive timelines, they wanted the application delivered less than a year. But from the requirement standpoint, they wanted to analyze geospatial data, very large amounts, in ways you don't typically see in BI tools. So freeform selection across any of the dimensions, the data set, including time series, extremely quickly. So our approach from the start was to take a tiered approach and we found that utilizing AWS services allowed us to accelerate the delivery across the board.

>> Tell about the tech stack and how did you mix and match AWS? What did you actually stand up? What were the key elements?

Ashley Videtto

>> Absolutely. So we started from the entire data platform looking at a tiered approach, and this is where the AWS services really helped us. So from the compute and transformation layer, we utilize serverless architecture, Lambda and Fargate orchestrated with AWS step functions to really deliver that analytics presentation layer to the specifications that Andrea and team wanted. But after that, what we realized is we could drive a lot of our ROI and time to market by utilizing many of the other AWS services. So from the storage layer upstream, we were able to utilize more traditional compute object storage to actually deliver that ROI for the customer. But as we moved upstream and really required the performance that Andrea and her team identified, we realized we could use a lot of the newer services, the faster storage compute. So things like AWS, FSx for Lustre, so we could have extremely fast shared storage and EC2 instance store to actually utilize the SSDs on the device. And further, when we actually delivered the solution, so it's a GPU-based solution, which we can get into if you would like, we were able to utilize ECS, AWS ECS to actually scale and deliver the containerized GPUs extremely quickly. So it was a win-win across the board.

>> Definitely want to get into some of the SageMaker stuff. I want to just talk about the geospatial stuff because that highlights the multiple data sets, got time series, all kinds of different databases. This is one of the beautiful things about the cloud is that you can take these data sets. How did you guys solve that problem with Amazon?

Ashley Videtto

>> Yeah, so not only was it a lot of data sets, these data sets were not joined in traditional fashion, so the dimensions weren't something that you could just join a primary key and foreign key. One of the requirements Andrea and our team gave us is that if I select any area on a map, whether it's millions of records in 10 different data sets, I want to be able to derive analytics from those. So the first thing we had to do was actually transform and make sure that we had a standardized geospatial data set so that all of the elements were in the same reference system. The oil and gas industry used very specialized reference systems based upon where they exist. And then, in terms of performance generating those type of exploratory analytics, since we were not able to identify a lot of your dimensions that you would build your aggregate indexes on, we had to make sure that we utilized a solution that could perform those aggregations extremely fast, which led us to a GPU-based solution. Then scaling and providing the fault tolerance and resiliency that you would find in these systems extremely difficult, but luckily we were able to utilize a lot of AWS services such as ECS to actually solve those problems.

>> So Amazon's stocking up on those GPUs helped out beautifully.

Ashley Videtto

>> It definitely did one of the requirements at launch, the SAP team wanted to make sure that it could support a hundred concurrent users, and GPUs are not known for their concurrency, so we had to spin off a lot of accelerated instances at the start.

>> Nice. And just so on your side, as this is going down, I mean, it's a high bar to put those standards out there, what was your thoughts?

Andrea Anderson

>> Yeah, I mean, it was certainly something we knew we were sort of reaching for. We knew what the ultimate outcome we wanted, but we also recognized that there was a big gap between, one, what we were delivering and how we were delivering at that time and where we were trying to go. Working with EPAM was great because we trusted them as a leading cloud and solution architect provider, but we really weren't sure how it was also going to come out. And very quickly, again, it was an accelerated timeline, but very quickly we were able to start seeing tangible ways to accelerate the use of the data and a product that we knew we could launch successfully.

>> Describe the difference between where you were and where you were reaching. It sounds like you were putting a bar pretty high. Tell us the scope of where you were here and where it landed. What's the difference?

Andrea Anderson

>> Yeah, so before a product like this, again, it was really about delivering data to clients and kind of saying, "You have the data, do what you will with it." This was about how can we take the workflows that they're doing after they've consumed the data, after we've provided it to them, and putting it in an application, a cloud-based application that we control that they can ultimately access this information. And we can better understand their workflows, their use cases, and ultimately starts creating more of that derived data, more of those machine learning models because we understand what they're trying to predict and what they're trying to prescribe for the future. So it gave us a better understanding of their use, and again, it kept them kind of within our systems utilizing more and more of our content.

>> What was some of the early indicators from your customers that this is working?

Andrea Anderson

>> It definitely, yeah, I mean, the retention was certainly improved. Obviously, revenue, kind of the standard things that you're looking for.

>> Of course.

Andrea Anderson

>> We expanded market reach. We found we were working with a lot more kind of financial services clients and business development who now found our data more accessible through this delivery tool and ultimately more kind of collaboration right between our-

>> So you guys are just making a product consumable and obviously, rings the cash register. That's good too.

Andrea Anderson

>> Always.

>> All right, so now on the technical challenge, what were the big technical hurdles? What are you most proud of? You get to point at the tech stack and say, "Okay, wow, this is really cool." What were the key elements?

Ashley Videtto

>> Definitely. So as Andrea mentioned when she was leading to some of the ML models and the implementation, so the customer didn't just need traditional analytics where you could optimize on the back end, but what the customer is asking for is the ability to generate those analytics on the fly. So create dimensions with calculations within the system across any subset of the data at will. This also included bringing the customer's own data. So we can get into the tech stack of some of the MLOps if you would like, but the customer has a lot of their own data, which sometimes, Andrea mentioned, is actually more realistic to perform the operations on than their own. So we had to bring that data into the system and then run the customized models on that data set. Yeah, there's quite a lot of technical challenges I could dive deep into. Accelerated timeline.

>> I think, well, there's a couple of things going on. One is getting the cloud element in there with the data, building a platform that's analytics-heavy in terms of capabilities that's easy to use is always hard to do. We know how hard it is. So I'm always interested in how that comes together. But then it comes down to, okay, we're now in a modern era where turning around product has to be faster, and I'm sure you guys have a lot more going on at the S&P saying, "Hey, we're going to start being in the product business. We got data and we've got products. How do we architect this so we don't foreclose the future opportunities?"

Ashley Videtto

>> Absolutely. So one of the key aspects of S&P is they have brilliant data scientists working on cutting-edge machine learning models that they needed a way to deliver those to the clients fast, especially in the accelerated timelines that the business was giving us. So one of the areas that we used to accelerate that delivery was we migrated the entire MLOps delivery into AWS SageMaker. This allowed us to use the inference and library toolkits to remove a lot of the boilerplate code so that the data scientists and engineers spent their time just developing the solution. On top of that, I mentioned how many users that had to support at launch. We needed to make sure that regardless of which model they hit, that the solution could scale quickly in real-time inference at a moment's notice. So we utilized SageMaker multi-model endpoints and set up scalability rules to ensure that as user requests increased, we automatically scaled to compute to meet that demand.

>> Talk about SageMaker. We've been covering SageMaker since it came out with one of the fastest growing products, but a lot's changed with the generative AI wave. How has SageMaker evolved from a practical standpoint for the folks watching out there, maybe have used SageMaker, heard about it, know that it's the central piece of their AI puzzle?

Ashley Videtto

>> Absolutely. So one of the key aspects, we know that this is a living, breathing product. We know the customers always going to want to bring in more data, have more functional requirements, and have more application capabilities, and the data scientists are always going to want new models, different types of models as Gen AI, as you mentioned. So we really believe that building on a foundation of AWS services will allow us to adapt and scale in the future as those needs arise. And mentioning Gen AI, all of the new capabilities with the bedrock and generating the agents and having that gated garden to protect the actual data and assets of S&P, is something we're really excited to continue delivering models on.

>> So Andrea, I want to ask you a question because I think what this brings back on the business benefit is that if you have now this capability with agents coming, I'm sure you now got a pipeline, more data from your side. You got all these data scientists. You got a great data marketplace, basically. You're data rich at S&P. You've got plenty of data so you can bring new products, new content in real time with agents under the covers.

Andrea Anderson

>> Exactly. I mean, one of the key pieces was really accelerating our time to market. So the ability for us to continue to grow. And so that was one of the key functionalities that we wanted was something that we could leverage so that we could continue to build without having to start from scratch every time. And so that's something that, as you said, we have a lot of data scientists constantly delivering and this is a tool that allows us to do that.

>> So you had a great part. These guys were good for you guys?

Andrea Anderson

>> Yes, absolutely.

>> And Amazon was strong?

Ashley Videtto

>> Amazon was amazing to work with and S&P Global were wonderful partners as well. It's rare to find a client that knows exactly what they want with all the requirements from the get-go, and Andrea and her team are excellent.

>> Well, Andrea, you're living the future. You get a glimpse of what was coming with AI and future data is, data is now the product, not just an ingredient. And you're in the product business now with data. Congratulations.

Andrea Anderson

>> Thank you. Good to be here.

>> Ringing the cash register up pretty good, right? I'll see you next time. Thanks for coming on theCUBE.

Ashley Videtto

>> Thank you so much .

>> We're in New York City. I'm John Furrier. You're watching theCUBE. We're here in New York for Tech Week and AWS Financial Symposium 2024. Be right back.

Andrea Anderson | S&P Global & Ashley Videtto | EPAM

>> The AWS Financial Services Symposium, presented by theCUBE.

Andrea Anderson

>> Absolutely.

>> Ashley Videtto, Solutions Architect at EPAM. You guys did a great solution. Thanks for coming on.

Ashley Videtto

>> Thank you.

Andrea Anderson

>> The Energy Studio is what that's called, the product?

Andrea Anderson

>> Yes, exactly.

>> So what was it? Clean sheet of paper, you have all this data laying around and say, "Hey, how do we roll it quickly to make better decisions?"

Andrea Anderson

>> What was the demand driver on your side? Was it more of the deals? Was it more we have this data? What was the key driver?

Andrea Anderson

>> Ashley, you're solution architect filling in for your colleague. Thanks for coming on. I appreciate that.

Ashley Videtto

>> Thanks.

Ashley Videtto

>> Tell about the tech stack and how did you mix and match AWS? What did you actually stand up? What were the key elements?

Ashley Videtto

>> So Amazon's stocking up on those GPUs helped out beautifully.

Ashley Videtto

>> Nice. And just so on your side, as this is going down, I mean, it's a high bar to put those standards out there, what was your thoughts?

Andrea Anderson

>> What was some of the early indicators from your customers that this is working?

Andrea Anderson

>> It definitely, yeah, I mean, the retention was certainly improved. Obviously, revenue, kind of the standard things that you're looking for.

>> Of course.

Andrea Anderson

>> So you guys are just making a product consumable and obviously, rings the cash register. That's good too.

Andrea Anderson

>> Always.

Ashley Videtto

Andrea Anderson

>> So you had a great part. These guys were good for you guys?

Andrea Anderson

>> Yes, absolutely.

>> And Amazon was strong?

Ashley Videtto

Andrea Anderson

>> Thank you. Good to be here.

>> Ringing the cash register up pretty good, right? I'll see you next time. Thanks for coming on theCUBE.

Ashley Videtto

>> Thank you so much .

>> We're in New York City. I'm John Furrier. You're watching theCUBE. We're here in New York for Tech Week and AWS Financial Symposium 2024. Be right back.