Name: New frontiers in visual data science & AI for the energy sector
Uploaded: 2025-07-22T17:03:54.600Z
Duration: 46 min 8 s
Description: New frontiers in visual data science & AI for the energy sector

Transcript for "New frontiers in visual data science & AI for the energy sector": Hi. Welcome. Welcome. So we're going to kick off with new frontiers in Visual Data Science AI in the energy sector. And there's a really a lot of action going on in all those three areas. So just to take a quick look at what's going on in the energy sector, pretty amazing that the fourth floating production storage offloading boat ship arrived in Guyana recently and now they're up predicting million barrels of oil a day. That's $60 a barrel, as they say in England, not too shabby, pretty cool. And the folks who are running that Exxon, we've got some Exxon folks in the crowd, right? We've got some Exxon folks there. I saw some Hess folks show up. So this is a joint venture, Exxon, Hess, CNOOC and lots of M and A. So last year, when we did this conference, it was Exxon Pioneer. We got Diamondback, Endeavor, Conoco, Marathon. There's a lot of M and A and all those customers have all been spot by our users. So we were kind of right in the thick of it and I'm really excited to see that the AI world is landing in the energy sector now as well. So I was at the SPE conference last year in New Orleans and SPE announced their foundation model for a large language model for the energy sector. And our group, Collide, everybody knows the Digital Wildcatters, Collide and Energy Bites. We're thrilled to have Bobby from Energy Bites on a panel this afternoon. They just got funding for LM in the energy sector. So lots of action in the energy sector and a lot of action in Spotfire, in Visual Data Science as we build this platform that Stephen referred to, we're doubling down on the energy sector, engineers, scientists, tough problems, combining immersive visualization with data science and energy sector knowledge. That's the three things that we're focused on to bring forward this visual data science platform that Spotfire always has been and that is re focused for us to make that a reality again in the energy sector. So what does that comprise? Well, it's this Visual Data Science and Rapid Application Development all in one. Visual and Geospatial Analytics with an immersive UX, drill up, drill down, drill sideways, Spotfire in the data. Data science, embedded, embedded R, Python, statistical engines that are running algorithms interactively when you do a marking. You can run an algorithm and land the results right on the map chart. It's a very immersive interactive experience for visual and data science analysis. Data management, by that I mean all the data, in memory, in database, on demand, cached, federated, virtualized. Every way you can consume data, we've got fifth generation hybrid in memory data engine that will run big data in memory on markings interactively to run visual analytics and data science. Really nothing else like it out there. Artificial Intelligence, we've been investing in that since 2018 and we've got some nice updates to show you today with our Spotfire co pilot, generative AI. And then application development and deployment, the other side of the coin here. It's not, couldn't be much easier than after you build a Spotfire application, right click save as to library, got a URL, send it out to hundreds, thousands of people to use interactively in the web player for daily decision making. It's about the simplest deployment mechanism I think I've ever seen, right? Flicks say those two library, right? So, and then with the energy sector focus that we have and we've redoubled down on that with Stephen showing up, recognizing the strength we have in science, engineering and energy sector, this is where we're really betting our future on. And you can see here this interactive map, right? So I'm looking at this self organizing map representation of seismic data there on the left, making a little marking. You can see the principal component weights at the top. I make the marking and I see from the seismic data, I see the rock where the hydrocarbon is. That's what I mean by interactive visual data science. You're making markings, you're running algorithms, you're doing visualizations to find structure, you know, in the data. Okay. So the Spotfire personas that we that we build for, right? The Visual Data Sciences scientist is someone who make sure this is clicker, but yes. Is someone who points and clicks, visuals, data functions into an application, saves it to the library for the end user, business user to come in a browser and have that same experience that they have building the application. So the visual data scientist is our real persona that we target things for, the person who understands all the ins and outs of the options in Spotfire, but it's all point and click pretty much, but how do you kind of make an application for a business user? Now the third persona is the developer. And this is someone who will write some Python code to create a data function, to put in the palette of the Visual Data Scientist or create a new visualization to put in the palette of the Visual Data Scientist and these folks know how to code in Python typically, R, JavaScript using D3 libraries, things like that. Now the Copilot that we're going to be talking about Fairmount today helps all of these personas, right? So it helps the developer because as you'll see, it generates Python code. You can ask it from a chat interface, hey, create me this data function to do x y z. It'll create that. It'll stick it in Spotfire for you and allow you to be highly productive as a developer in Python. The Visual Data Scientist, you'll see that the Copilot helps with authoring and analysis. You can ask questions about the data, help you visualize certain data in certain ways as you build your application. And then for the end user, it will explain what you're looking at. It'll interpret it. It might create a report for you. So the Copilot works, in the analyst, in the web player, and allows you to go through those, sets of personas. Okay. So what about the energy sector? Well, there's a lot of personas and use cases in the energy sector. Right? So I know a lot of the crowd here is gonna be upstream. Quick show of hands who works in upstream. Quick show of hands. Yes. I figured that. Well, most of the audience here. I did see some midstream folks show up as well. But basically the reservoir, a geoscience engineer, the drilling completions engineer, the production engineer, as Steven said, these are our key personas that we want to make happy and successful. But I did see a few folks show up from downstream trading, midstream. I saw some please raise your hands to folks at the back door. I know I just spoke to you. But yes, a bit of a smaller crowd in the midstream, but we also have use cases, you know, and personas that we work with there. We're trying to do more in that in that area. Okay. So this year, as Steven mentioned, we've been working hard on this Visual Data Science platform, taking input from customers, understanding use cases, visualizations, data science functions, and we've been field testing those functionalities and we've been hard at work building them now into our product as supported applications. We've been hardening the Visualization Data Functions for product level, security, scalability and support. And this is our chance today to show you what we've been doing and close that loop. So as we've taken our inspiration for customers, now we're going to deliver that back to you and you'll see what we'll be working on today. And we're really interested in your feedback, for improvements or what do you like, what you don't like. And you'll see today that we're creating and sharing these sort of visual, data science components and tools for building applications, and that's what we want to get your feedback on. So don't be shy at the end of the day when we ask for the feedback. Alright. So let's now jump into what are we talking about with visual data science, AI, energy tools and applications. Here's some examples that are on our demo gallery on our on our website. Drilling, trading, production, and completions, data replay of live data, process engineering applications. And these have all the elements of Spotfire componentry I've been talking about inside of them, being built fit for purpose on those certain use cases. And the way that works is you got to load data, explore it, clean it, build visual data science functions, refine the UX, deploy it. That's how you create those applications. And I know a lot of the companies here today have got teams that do that, but it all boils down to how do what what connectors do I have, what visualizations do I use, what data science functions do I create, and this new category of tools called actions that we're going to show you today, which allows you to combine the configurations that you might do for a a data function or for a visualization into a single click that you can save and share, across the library to other folks and then getting that out into production. And again, the Copilot is helping in each of those areas. So I've been talking quite a bit about this Copilot. I should just, you know, get it off the table and run a quick demo for you. Okay. So I've got some wells here, green producing wells, gray non producing wells. I've got production data, and I'm basically going to, tell the Copilot bring up the Copilot. You see the Copilot's got some suggested questions to begin with, but I'm going to create my own question here. Create your data function to read the LAS file from a URL, bring in the data, smooth the gamma ray, calculate the B shale, put out a long skinny format of that data, put out a wide format of that data. And so what that copilot does, it actually creates the code. It sticks it inside of the data function experience that you have in Spotfire. It creates the parameters, it maps the parameters, the comp the code is all commented, pick it up, you run it, and then when you run it, it respects all of the flow. So you've got now a validated flow from API to well ID to the reading files, putting out the data, smoothing the gamma ray. That all happens for you and it's all tracked inside of the Spotfire data canvas. So I have a history of what I've done there. So now I'd say, okay. I've done that. So Copilot, what should I be looking for in the data that I've just brought in Spotfire? So the Copilot comes back and it says, well, let me just sorry to pause this video for a minute. It's suggesting, to look at those data in a well log analysis. It's suggesting to compare the well log to the production data, and it say, here's the correlations and reservoir quality using the log data, how do I understand the relationship between that and production. So I go to my panel, see the panel now. Let me actually pause that for a minute because this is new, and you'll see in the what's new section this this kind of stuff, but we now have a fly app where you can pull the well log and other specialized energy visualizations are now available to to run, you know, inside of the you know, from the menu system. Okay. So now I do a well log analysis on that LIS file. I can see the profile of the gamma ray. It's smooth per the instructions of the Copilot. You can see this certain region, a boronoid polygon, a region of wells, and I can in the surrounding that, I can browse the well log analysis to look at the gamma ray, and I can line that up with the producing well in the middle, to see the production as it relates. So here's the production well, mark that one and look at the production over time. So I'm trying to combine production with well log analysis for a certain region and that's happening sort of point and click in conjunction with what the Copilot has done for me in setting up that data. So I can kind of cruise around and look at the well logs and look at the production, of the wells in that region. Okay. So that's what we're trying to do with the Copilot. And the Copilot, you get instant value from the data, you can create and modify visualizations, you can interrogate the data for immediate insights, obtain analysis suggestions as click buttons, create and execute data functions, and also you can upload your own documents and have that as context in the Q and A that you're doing with the Copilot. So it's not just a Copilot, the Spot by helper, it's also context as provided by you, the customer, to bring your documents into that experience. Okay. So we're now going to go through those steps from, connecting to data, visualizing it, predicting data science functions and deploying an application. I'm going to show you the components that we've been working on as we've been going along that journey and remind you of what we have in Spotfire. That's part of the big reason for this this presentation. So when I say we read all data, so we bring data into memory through information services. That's a point click query generator. We also do direct queries that will run aggregations inside the database and bring the aggregations back into Spotfire. That's what I mean by hybrid in memory. We do direct connections through ODBC, JBC. We read in energy sector specific files. We read in parquet files, CSV, other things in text. And we have advanced data services throughout Spotfire Virtualization product. That's a separate abstraction layer with virtual views that federate data from multiple source systems. And then we do caching, so we can cache the info links to the Spotfire server. We can cache the views from the data virtualization. We can schedule updates into the web player. We've got SPDF binary data that you put in the library, for very fast performance of cached, Spotfire data. So caching is the important part of performance. And if something doesn't change in a long time, you should just cache it. If it's changing a lot, go get it. But that kind of hybrid approach really improves performance. And so we've got this hybrid in memory compute with approximately 300 different data sources. You'll hear a lot about different combinations of those data sources with Spotfire today. Now just a little moment. So partnering with so fast though, in this fifth generation engine, we've been at this for twenty five, thirty years. This fifth generation Hybrid in Memory data engine allows you to do data on demand. So any of those data sources, you can have data in memory, but then you can go get data and bring it in as well. So if you find an area when you spot the fire, hey, I want more data there, just go get it and swap it into the in memory to do the performance. And that can come from any source we can do on demand. This is part of the reason that big data in memory, in database combined with data on demand is such a powerful construct for analyzing lots of data. So just a little sidebar on Advanced Data Services, data virtualization. We're customers who use this get so much value from it. There's extra data access layer where you can control metadata, bring data in from multiple sources. You can create a catalog of views. That catalog of views shows up directly inside of Spotfire that you can build your analysis on top of. It provisions data without movement, allows you to find, understand and secure and consume data from one place, optimizes the query, it's got intelligent caching. It determines if something hasn't changed for a while, I'm just going to cache it. So it has a really smart way of dealing with data that's changing versus data that's not changing and I say natively integrated with Spotfire. And so our vision here is that we can bring subsurface geology, petrophysics data, engineering data, operations data, real time SCADA data, financial data, syndicated data. You can have all those data sources and have this layer, a virtualization layer that brings some of those together that allow you to build Spotfire applications for GNG, reservoir to production, like I was showing a minute ago, completions, drilling to production, asset valuation, competitor insights. These are all use cases that can be built upon a collection of virtual views against your multiple source systems, giving the end user more control and IT less headaches, that extra virtual view virtual layout. Okay. So that's the data part, the data connectors part. That's the first thing you got to connect the data. So what about visualizations? So we created some years ago this notion of extending Spotfire with JavaScript using often D3 as a starting point. But we've built a bunch of visualizations over the years in this visual mods or modules area. Some of those are energy specific, you'll hear about them today. Others we're just building because someone of our key customers says, I'd be great if we had this chart. So we'll help them build that chart. And so we've done a lot of those. As Stephen mentioned, we're really doubling down on energy sector as a key market for us, and so we've been hard at work building the well log analysis I spoke about, the well bore analysis, gun barrel plot. This three d surface and line chart has proved to be incredibly useful in a bunch of different use cases. Brad and others are going to show off and Athia are going to show that off today in a number of different contexts. But these are some of the specialized visualizations we've added into the product for the energy sector. In terms of the community, excited to get the Marimekko plot out there. Some of you folks might know that Spotfire began its life at the University of Maryland Human Computer Interaction Lab with Ben Schneiderman, who runs that program. This is a visualization that Ben invented. He invented the tree map. He's got different versions of the tree map, 10 of them inducted into the Museum of Modern Art in New York City for the analysis of various data sets. This is one of his specialties. This visualization doesn't do it justice, but we've now got that one into the community site. The rose chart, very good for looking at directional data. Wind energy has been de emphasized this year, but it's really great for that. But it's great for anything directional. These process maps, you'll find them very useful for documenting processes and mapping out a data journey that you have. I included the spider chart up here because this is one that's very popular in the energy sector. You can look at different types of artificial lift systems, their production, really compare on in this case, of course, it's eight dimensions. You can see which wells are doing well in certain scenarios and comparing them to others. And then just a personal favorite that we're working on, anybody heard of the band Joy Division? Quick show of heads. Anyway, this is the cover of the Unknown Plutus album. This is sound waves coming through space and a graph of it. And anyway, this is called the Joy plot as a tribute to Joy Division. But we've got a whole community of visualizations that I'm sure you'll be able to bring into your analysis with great effect. And then, of course, streaming data visualization. So we have all of the charts, the graphs in Spotfire are capable of being updated in real time. And so real time data, all of the visualizations. I've seen other products out there where they'll update a line plot or time series chart over time. We update every visualization over time. So you see the map is updating. The bar chart is updating. The pie chart the lines of the line plot, there's basically a live data mart inside of Spotfire that it just picks a row off at a time and updates the visualization, very useful for real time drilling, real time completions, monitoring production on a bunch of wells around the world. And so the combination of this real time streaming biz with data at rest and integrating that is particularly useful in a number of use cases. And you can configure it to get alerts and notifications on certain equipment that might not be performing well. So it's a nice adjunct to the whole concept of visualization inside of Spotfire. Okay. So I've talked about connectors and visualizations. Next part of the puzzle is data functions and we've been hard at work creating a whole bunch of those, especially in the area of time series and geospatial, two areas for us that we've been working hard on. And excited about the fact that we also, some years ago, acquired this product called Statistica and this has a slew of algorithms. This is a SAS competitor. It's got thirty years of algorithms in it, everything from inferential statistics, hypothesis testing, classification, regression, predictive modeling, quality control. And we've now come up with a process we'll talk about in what's next session at the end of the day about bringing these as data functions into Spotfire to greatly enhance the collection of data science functions that we have inside the product. Now these data functions, as I mentioned, calculations powered by Python, R, Tera, Statistica and we've built that in this product, that's new. Or you can write your own data function, haven't taken that away, or you can download from the community. You can manage these from the fly out. So this is the fly out of the data functions. This little fly out here, these are the visualizations, these are the data functions, these are the actions. So we've been architecting the product to take these extensions for years, but here's a collection of data functions that now you can run and add to your analysis directly from the flyer. When you add it to your analysis, we automatically generate a user interface that you can point and click and run it. But you can also run it on a marking when you make a WASSU on a map chart, for example, you can run one of these data functions. And then the deployment for the administrative folks in the audience, Spotfire nodes and services architecture, pretty simple to manage the packages, governance, licensing at a very fine scale. You can decide who gets what as an administrator at a very controllable manner. And then with the Copilot helping you code gen to create your own data functions, that's a big, big uplift. Okay. So I'm gonna talk about just a couple of sections of those data functions. I'm gonna talk about time series analysis and geospatial analysis. So analytics in time. You know, in the energy sector, that's a lot of equipment. Sensor data, skated data, it's a lot of it. How do you make sense of it? I've added a smooth curve on some of this stuff to help make sense of it, so we're doing quite a bit of that. But you can also down sample it or up sample it. Just gotta deal with this massive collection. You're trying to get insights in it. So I've chosen an example here to get your participation to help me understand what are these curves over time. Okay. That's a bit too generic. Let me tell you that this is power usage in buildings, and let me tell you that the x axis is Friday, Saturday, Sunday morning. Okay. So what type of building do you think is in the top left? Any ideas? It's a business. It's a weekly, not much usage in the evenings, but during the day, a lot of power consumption. So these are businesses' weekly everyday operations. What do you think this one is? Oh, I already told you. You can see this one is only getting a lot of usage at night. So these are probably nightclubs. Right? These are probably nightclubs. You see the usage on Friday night, Saturday night especially. Okay. What about the next one? Single family homes. This is kind of pretty consistent, right? This is just consistent across the weekend. Families don't go out a lot a little bit, but, okay. I'm going to skip the next one. Let's go to a tough one. What about this one here? What do you think that is? Well, that's a downtown business, right? Friday is pretty busy. Saturday, a little bit in the morning. Sunday, not so much. And then the last one, these are condos, so it's not a lot of usage on Friday, but people show up on Friday night, they hang out the weekend. And this one at the top here, I deliberately skipped that. That's probably a broken meter, but any case, you can see what we're trying to do with time series data. We're trying to smooth it. We're trying to sample it. We're trying to create patterns and insights. And so we've been working hard on smoothing. So we've got a several number of smoothers that can smooth that data, working hard on resampling. A lot of times you get data point every second, you don't really need that. So how do you downsample that to just what you need, maybe a daily reading or maybe you can use the internal correlation structure to downsample. But we've been doing quite a lot of work on time series. One other thing that I fear is gonna show in the next session this morning is this thing called dynamic time warping. And this is not something out of the Rocky Horror Picture Show. It's not let's do the time warp. It's not that. It's, you've got similar patterns in time and you're trying to they misalign. So how do you join them up? And so this warping thing aligns those sequences. It's a one to many mapping. This is a bit washed out in the screenshot, but let me run this. So you can see that you can warp, you can move with this one to many mapping, those sequences to be in a line, and you also get a measure of similarity, out of the analysis. It's helpful. You'll see our Pia's got a great demo on this coming up. So that's what we do there in time, but we've also decided to flip this and turn it into depth. So you can now walk your depth profiles and that if you've got a currently drilled well that you have gone in blue, a previously drilled well, you can kind of match these up and learn from a well that you've already drilled and apply that knowledge to the new well that you're about to drill. So again, Appiah has got a great demo of this, and he'll show that to you. He also wrote a terrific article on our community site about how time warping applies to the oil and gas market. So you'll see more of that today. And then finally, on the event stream, time series analysis on the event stream. So you can see with data from completions or drilling, intelligent equipment production, you're getting time series data from the equipment. And so this is a production this is an ESP. You can see the ESP in here. Pressure is starting to go up, current is coming down, that's a problem. That could mean a block tube or you're about to go into a non productive state. You're about to have some downtime, which you don't want to have. So the ability to do calculations directly on the event stream helps with condition based maintenance, predictive maintenance type things to keep your equipment running. Okay. So there's a few examples of what we're doing in time. Now let's move to space and see what we're doing now. So we've been super busy at looking at geospatial analytics, whether we want to do spatially join, transform coordinate systems, make choropleth maps, do wrangling spatial search. So quite a lot in the top section here, we've been doing quite a lot in location analysis basic tools. Also in the bottom section, site selection, spatial DSML, interpolation, contour analysis, isochrones. We've got a specialist here from the geospatial energy sector, Amir. You want to stand up real quick? And in the back there, he's also next to Athir, the guys doing all the time series work. So those guys, Amir and Athir, in the back there. If You got any questions on time or space? Go talk to those guys. They're not Star Trek people in in running around the universe. They're actually spot by people who know all about time and space. So take your questions for them on break. But the beautiful thing about Spotfire is this multi layer map design that we've got. We got this many years ago from an acquisition. It's second to none in the industry. It allows you to put down a map layer, to put down feature layers on top of that, additional feature layers, put marking layer on top of that. And we can calculate within and between layers various things. We can calculate, okay, within a polygon, what is my average production? What's my type curve? Or if I've got geology data on one layer and I've got production data on the other layer, how is what geology variables are predicting production? So we can do and we can do that all on a market. You can just go and mark a section of the map, you can run a data science function and get some pretty deep inference on the spatial level from those analyses. We can bring in satellite data, aerial photography, geology, different coordinate systems, tile mapping services, web mapping services, Esri shape files, Esri layers that come in as feature layers, GeoJSON layers, real time geospatial distributions. It's really second to none the sophistication of this geospatially led analysis, which is hugely important for the energy sector. And you'll see that we've now allowed you to put all kinds of maps symbols besides just circles and so on on the map that help you with that. And so we've been busy with doing spatial joins, points and polygons, geodetic polygons, contours, kernel density, proximity, nearest neighbor, CRS transform, geospatial data streams. These are all examples of what we're working on, but we've got a long list of functions and we're curious to understand what you want that's extra functionality in this area. Okay. So we've done connected to data, we've visualized it, we've done some data science on it. This final section actions that Brad and Matthew are going to go into in gory detail allows you to configure a visualization or data function to make a UX experience that's simple rather than having to click a whole bunch of buttons. We can save that in script and we can save and share that across another other analysis in the library, and this, uses our JavaScript API. So that's the third little area in the fly out is these is these, actions that we've been building, and you can build those yourself. Like I said, Brad and up here are going to go into that in great detail. But examples are how to configure the well logs, so we move outliers, calculate new columns, configure the display, find similar wells, calculate a production. So things that are like macros that you want to do multiple things and multiple configurations in Spotfire, you can join all those together and have a reusable action saved to the library that other people can can bring into their analysis. Okay. So bringing that all together from connectors to visualizations to data science functions and actions, let's have a couple examples now of assembling all those pieces, to create, you know, a workflow. Okay. So I'm gonna give three examples, one from reservoir characterization, one from drilling and one from production, just to give you a flavor for it. Okay. So this is our new well log analysis. These are two wells, well A, well B, in the Gulf Of Mexico. You see the production curves that go with these trellis displays and deviation survey at the bottom. Again, my colleagues are going to go through this in great detail, but the wells on the right hand side of the trellis display of two wells includes the formation tops here. So this is the Wilcox one, two, three and four. It includes the completion interval, which is this gray bar here, which shows where the actual completion occurred relating to these parts that you've been in the drilling completions area. And then it includes a bunch of data on the panels. So this is the gamma ray here. Got this nice shading. It's a gradient fill, but here's the gamma ray. This is the lithology variable, so silt, clay, sand type things. This is the porosity and volume of water. We've blown that up here because it's so important and the green is colored indicating the hydrocarbons, the blue indicating the water. And these little points, these are little points of pressure that were measured. A lot of these things are new from what you saw last year, the ability to display these data in this compelling way. And then the next one is the neutron porosity and the rho b. And again, where those curves cross over indicates whether it's sand or hydrocarbon, lined up with the gamma ray analysis and lined up with the porosity analysis. And then this is the resistivity in the final plot. So, let's see, did I have to run this video from here, I forgot to run this. And the data provided by Joe, it's PSI, it's really awesome data. This is the Petrophysical Solutions, Inc. And the richness of what we've been able to get from the petrophysics variables in this analysis goes way beyond just a single gamma ray. It gives you a lot of depth in the richness in the interpretation. So in addition to those two panels there for the two different wells, we can start playing with the path. So here's the deviation survey. As I rotate that around, you can see that well A covers much more of the Wilcox hydrocarbon producing rock than Well B, which sort of gets a little bit at the end. And so I'm going to line these axes up with on true vertical depth, and then I'm going to highlight one of these sections here. This is the Wilcox B, and I'm going to tag that. And now I'm going to show that in the second section and now you can see communicate with your team about Wilcox 2 in the two drilling operations they covered more of that territory in Well A and less in Well B. It gives you that overall interpretation of how production was bigger in Well A because of where you drilled and how the deviation survey helps you with that. So great data set from PSI and this is just a really high level snapshot of it. Theo is going to go into a gory detail with this about how we can configure that and then the new features that we have in there. Okay. The second one example is from drilling, and this is a real time example. And here I've got in this analysis, I've got across the top rate of penetration, weight on bid, the hook load. These ones here, the rate of penetration, weight on bit block height, the torque, the hook load, the RPM, here's the drill state. So what is the state of office? Is it rotating? Is it tripping in? Is it tripping out? Is it, what's the status of the drill? This is real time data coming off the drill bit. So this is WIDSO type data. On the right hand side, the brown curve is the expected drill path and the blue curve is the actual real time path. So you can see we're veering off a little bit, and so how do we bring that back on track, which we do have algorithms for. One of these algorithms that I'm mentioning, rate of penetration and weight on bit, you can derive bitware from that, real time bitware estimation. This gap here, you can use those equations for bringing that back onto the original path. So a lot you can do in the operations from a drilling operation by looking at your data in Spotfire. Okay. So final example is on production surveillance. So here's a collection of wells, and what I'm doing at the top here, I'm changing those alert conditions. So I say, okay, if the pressure goes above this number, I want the to be alerted. And so you can see the color now changes to red when a well goes above that. So when you raise the threshold, then you're gonna get less alerts And the other one is the motor temp and the pressure. I'm arranging that I'm moving those thresholds around for alerting. So the alerting comes when the color gets above a certain goes to red and when the size of the point gets big. So you can tell a big red dot means both of these are in trouble on those two analyses, pressure and mode temperature. And so I can monitor these wells across the region. You can see that the lines change color when they hit the various thresholds. I've got a table here of alerts that I can hand off to a tech folk, person who's in the region who can go and look at the wells and they can compare this condition to the predictive maintenance and they can decide how to do all the repair work they need at once and this is in avoid non productive time. Okay. So that's, the three examples to show how we bring together connections, visualizations, data science functions, together into an application. I'm gonna wrap it up with, the Copilot. So how does the Copilot help in a lot of this stuff? So the Copilot allows you to do general q and a, Spotfire q and a, have a custom conversation, interrogate your data, create charts automatically, generate code, and you can bring in your own documents. We've got all of our Spotfire documents in there. We've ingested all of our Docker's Spotify sessions, all our YouTube videos, all our product documentation. That's all been tokenized and is available to a query. We've got some special topic modules. We're adding to that different topics like completions, for example, and you can bring your own documents. We run on all clouds, Microsoft, AWS, Google, using their language models, OpenAI, Flow, Gemini, and that's how it's set up. Now we've been adding agents into this architecture through the MCP API, and so here's an example where we've got drilling reports. These are coming in as PDF documents and we've been able to parse those in the graph database and interrogate those to get inference from across the different reports. So you can ask a question that is like having all those PDFs in a database and just ask a question about the collection of those reports where you analyzing, what wells are producing more than others, that kind of thing. And so we've been making a number of case studies for these areas, whether it's an agent or whether it's coming directly from Copilot. We've got a use case for overall equipment efficiency, manufacturing yield, three d subsurface, completions, well log analysis. And then we've got this thing we're just calling AutoGen. This is new, hot off the press. We're trying to figure this out. But the way this is working is that we've had a long journey with Spot with AI and Spotify. We put recommendations in Spotify 10 back in 2018. We had interactive AI in 2020. Now this recommendations engine, it's a continuously running AI engine that's looking at all of your data and it's finding relationships, associations between variables. And when it finds one, depending on the variable type, it will suggest a graph. So here, for example, it's a scatter plot with two colors that's capturing a relationship between this variable and another variable. It's giving you the visual. It's a thumbnail. You click it into your analysis. So you can build pretty much your entire application right now from Spotfire recommendations. That was in 2018. And we've added the Spotfire Copilot version one and two and we've added agents. So this thing that I just briefly showed you there is about how do we take the recommendations engine, input about the associations and the suggested visualizations, how do we feed that to our latest version of Copilot and produce all the graphs and produce all the functions. We're trying to basically build analysis for you in one click. That's kind of what that example does and and, you know, it's again, it's a work in progress, but that's something that we're we've been working on here with this slurry example. So stay tuned for that. We're excited that Microsoft is recognized as we started with Microsoft as our partner in Azure, for the Copilot, before we got into AWS and Google. They've been funding us to get it up there, and they just wrote an article early part of this month about how that's available and the success we've had with using their Azure AI Foundry elements for OpenAI service and so on for running the Spotfire Copilot. Okay. So I'm going to wrap it up here. I think I'm kind of pretty much right on time and a little bit early actually. And just tip of the lid to our partners and the ecosystem that we have here. I know a lot of the folks who are in the room today, also the Kion guys, the Genware guys, Brian McDowell from Sabatas, kind of a wild crazy guy. Are you here, Brian? Probably out getting core samples in a truck and bringing them back home. We've got the guys from Enatel, Wiserock, S and P Global, Innovata, Petro dot ai. So yes, we couldn't be happier and proud of the ecosystem that we've built over the years and really want to acknowledge everybody in the room here. And also a big shout out to the SpotBot Data Science team, our Customer Success team, our Product teams, Engineering teams. It's been a big lift since Stephen arrived. We kind of refocused our business on the energy sector, our product and we've been all been working super hard to kind of bring this stuff together to show you, but today we're pretty happy about it. And finally, don't forget about the community that we built online. We started out a community site just last year, 2024. As Stephen mentioned, we spun out of TIBCO, we've got our own standalone business unit, we've now got our own website, our own community site. This community site is really active. I think, Elyse is here. She's big in the community site. I think we've doubled our activity in the last six months. Spotfire experts, customers, partners are all answering Spotfire questions on a daily basis. Our Spotfire program, Jose, you wanna put your hand up there? Jose runs Spotfire. He's gonna be on on the panel later today. Very strong monthly live hands on webinars, a steady release of Spotfire quick tip videos into our Spotfire YouTube channel. All the content on the community is very customer driven, So the content is chosen to be worked on in response to frequently asked questions, new product features, industry use cases. We're always open to new topics. If you want to be to create a certain Doctor. Spotify session on a live or a quick tip on the topic of interest, just please let Jose know. There's an email address, I think it's just spotfire.com and you can join in on the community. And you have to be a member of the community to take advantage of some of the new features that I showed today that require a community login and a community account to access. So we're working super hard on the community. spotfire.com was launched just before the community. Doctor. Spotfire has been around forever. But please use today as a way to give feedback to us on what we could do better in the community with the product and with the Spotfire program. And with that, I'm going to wrap it up and thank you for your attention and thank you for being part of this great ecosystem and community.