Name: Transforming high-tech manufacturing with Spotfire® Data Science
Uploaded: 2025-05-13T16:16:09.331Z
Duration: 55 min 47 s
Description: Transforming high-tech manufacturing with Spotfire® Data Science

Transcript for "Transforming high-tech manufacturing with Spotfire® Data Science": Hello, everyone, and welcome to today's webinar, transforming high-tech manufacturing with Spotfire data science. We're thrilled to have you with us. I'm JP Richard-Charman, and I'll be your host for this session today. Now before we get started, just wanted to cover a few housekeeping items to ensure you have the best experience. The webinar will last up to forty five minutes with a q and a section that will be held at the end. If you have any questions at all during the presentation, please don't hesitate to use the q and a panel located on the right side of your screen, and we'll address as many questions as we can during the q and a segment at the end. We've also made a few assets available linked to today's webinar in the doc section of our webinar platform. So that is located, in the same panel as the q and a, so right next to chat and q and a in that panel. So please do feel free to access these. After today's session, a recording of today's webinar will be made available on demand, and we'll also email you a link to that shortly after the event. Now with that, let's dive in. So I'm very excited to introduce our presenters today, Alessandro Chimera, our industry solution lead here at Spotfire, Dan Rope, our senior principal product manager at Spotfire, and Tomasz Jurczyk, our lead data scientist here at Spotfire. Now with that, I'll quickly go through the agenda of today's session, see what you can expect from today's session. So we'll be diving into why visual data science is important for industries, so specifically high-tech manufacturing in this case. We'll then go through give you an overview of Spotfire data science and those key capabilities that you can expect. And then we'll dive into a demo which Tomáš will do, really focusing on how to use Spotfire data science and its capabilities in high-tech manufacturing. And then we'll be finishing off, with some of our key success stories, so some of the successes our customers have had in high-tech manufacturing using Spotfire, and then we'll quickly move on to our q and a segment. Now with that, I'd like to hand it over to our first speaker, Alessandro. JP, thank you for the introduction. So I'm excited to kick off today's webinar session on how we can transform high-tech manufacturing analytics with Spotfire data science, specifically in the semiconductor manufacturing industry. So overall, the manufacturing industry is data intensive with data coming from sensors, equipment diagnostic, process logs, production metrics, and historical records. And staying ahead of the competition depends on one important factor, how fast and accurately we can turn data into useful actions. Unfortunately and too often, data are buried in data silos or waiting on other teams or stuck using tools that don't talk to each other. And so the question becomes, do your engineering decision makers have the tools they need? This is a fundamental question. Our factories and product teams are filled with smart people. But are we doing all we can to help them to pull insights out of the massive and messy data that are being created everywhere around the business? Do they have the tools that they need to understand this data? So today, we'd explore how Viso data science with Spotfire helps you to empower your teams and make key manufacturing decisions faster and with more confidence. I started this presentation by introducing the notion of Viso data science. But what is Viso data science? So high-tech manufacturers will always have, these two categories, of software that that we can see here on the on the slide. On the left and right, specialist tools and then also, statistics tools. In regards of the first category, the statistical tools, most engineers are not statistics experts. Today, we might call them data scientists, but, those folks specialize in knowing which algorithm they use and how to treat them for a specific purpose. But it's tough for engineers to leverage that knowledge directly. I hear pretty often, hey. Send me your data and I'd send back some answers. We can go over them, in the next Friday. And someone else problem is always in front of you in the line of the statistics office. And even today, it's not easy to put advanced models into practice in the business. You have to write some Python code to build application and then eventually write some JavaScript to put that in an online portal. And then when it comes down to specialist tools, the second category of software, you have to be a specialist to operate them. And there was always that one guy who knew every feature of the product. Most of us would just create the standard reports or walk through the standard processes. But it turns out that most of the problems in manufacturing are not really standard. It's not easy to extend specialist tools to incorporate other data that you need. They're just not, let's say, they're inflexible. But here, Spotfire is different. It is designed to be the indispensable data analysis technology for semiconductor engineers. And by combining agile exploratory visualizations with advanced manufacturing specific algorithms, Spotfire puts the data at your fingertips. And here, Spotfire helps you to bring together data from anywhere and display it however you like. So engineers can wrangle data, find and clean outliers, and create new variables and metrics on the fly. Spotfire so so natively supports Python, R, and our statistical product to create reusable data functions that you can just drop in and they become new tools in your menu. And this is what we call, this will data science. And here's an overview on how domain expert across the industry can benefit of a digital data science platform. So we have the equipment and process engineer. They can optimize and monitor equipment performance. These engineer can analyze data from sensors to identify issues, prevent downtime and optimize production parameters while applying their technology knowledge to fine tune processes. And then we have the yield and defect engineers. They can enhance yield analysis by rapidly identifying patterns and defect data across complex, multi stage manufacturing processes. And those engineers can then leverage their expertise to address root causes, reduce defect rates and maintain high yields. And then we have the product and test engineers. So with Spotfire, they can accelerate product testing and troubleshooting, enabling quick detection of inconsistencies and improving test efficiency. And those engineers can apply their domain knowledge to validate insights and ensure reliability of product quality. And then last but not least, our data scientists. They're empowered with a flexible platform to support all the teams by building data functions, developing Python or R code, and delivering prototypes. Often, it's really a teamwork. One output analysis of one team may be the input of another engineer or another team. We also cover a multitude of use cases from predictive maintenance analysis to a productive characterization where the visual data science approach delivers better data science improvements across all stages on the high-tech manufacturing industry. So overall, we deliver a platform that is not just powerful, it is built for how engineers think and solve problems. Spotfire combines agile interactive data exploration with advanced manufacturing analytics that allows your key engineers to use their expertise to ask and answer questions themselves and to use their creativity to come up with novel solutions to complex problems. Overall, we are widely deployed throughout the industry. In fact, the most important global semiconductor manufacturers, they use Spotfire throughout their operations. We also count some of the largest equipment vendors, that are our customers. They use Spotfire to develop their own product and develop reference processes that they share with their customers. You can even find Spotfire inside of some vertical solutions. And overall, we have been in the semiconductor industry for more than twenty years. And today today, we are just redoubling our focus on high-tech manufacturing to provide the best possible tools tailored to the industry. And then also, when we have time, I'll walk you through the few customer use cases, and let's see if they align with challenges that you might have also in your business. So later in this presentation, we are going to show you a demonstration of Spotfire. But before I want to highlight three important capabilities. So Spotfire is both human powered, putting the data at your fingertips so you can filter, link, and drill down, and also AI powered with a building recommendation engine to reveal relationships. We can handle all kinds of data, even multi layered spatial data like swabbing and defects where the spatial relationship between them that matters. And Spotfire automatically recalls our steps so that we can just save the file. And when we are done, we can have an application that we can use and share with others in a centralized way. You can also make edits to those workflows after to add more steps or change your approach. And finally, Spotfire can be extended in in other ways, including custom visualizations and custom automation tools and jobs. So and with that, I hand it over to Dan Rope who will dive deeper into specific Viso data science cover abilities of the platform. Alright. Thank you so much, Alessandro. Okay. Alright. So I wanna talk a bit about Spotfire data science today. But just to get into that a bit, I wanna first talk about our vision with Spotfire because we're very excited about this vision. So those of you that use Spotfire today to solve complex problems in high-tech manufacturing, you're well aware that it's that unique combination of analytics and visualizations that you get with Spotfire that really enables you to solve those problems in a way that no other product really can. That's an area that we're really gonna be focusing on and and doubling down on, that which really supports what Alessandro is referring to as visual data science. In addition to that, and what we're adding new now is a more of a focus on industry, bringing in some of these specific features for industry, whether they're algorithms or they're interactive features or they're visualizations or they're data sources, bringing more of those natively into the product for industries such as high-tech manufacturing, and and focusing on that. So it's more that you can have them available in the product versus having to go create them yourself or maybe download them from our community. Just bringing these things more natively into the product. And then retaining that enterprise scale, the ability to have a governed secure, Spotfire for thousands of users can access it and working with data of of all different kinds of sizes. So we feel these three things come together and really make Spotfire something unique, and that's what our that's what our vision is gonna be now and going into the future. So Spotfire data science. Spotfire data science, this is gonna be this builds on top of Spotfire analytics. This is the place where we're gonna have more capabilities for data understanding, more capabilities to do data preparation. This is the place we're gonna have features such as modeling and prediction that you can use for for working on your problems. And for high-tech manufacturing, we're gonna be adding specific visualizations, new algorithms, and connections to data sources that will be available to you, in in Spotfire data science. So let's drill into what of what some of these things actually are. So first off, we're adding new data function, new algorithms into Spotfire that'll be available out of the box for those that are are using Spotfire data science. Now our initial batch and there'll be many, many more to come. But our initial batch is focused on time series, geospatial, and working with missing data. And so if you're working with Spotfire data science, you have the ability to use these data functions, configure them, execute them, and use them to analyze your data. And they are largely chosen to focus on areas certain in energy, also high-tech and high-tech manufacturing, and some cross industry problems as well too. So I'll talk about a few of these. So time series. We have added a few, algorithms for working with time series data. There's smoothing and then there's also resampling. With smoothing, this allows you to, identify trends. A lot of times, machine data can be very messy. There's outlined values, and you wanna be able to see the trends or identify the outliers. And so this this data function will allow you to calculate smooth curves that can be applied directly to those lines, And you can see where those outline values are and what the underlying trends are. It's highly configurable. It offers a variety of techniques that you can use so you can, best get the fit towards your data, and even choose, you know, parameters to adjust that smoothing level as well too. Resampling is useful anytime that you need evenly spaced time, time intervals with your data. That's often needed to prepare for downstream analysis. And it allows you to adjust the frequency of your time series too. Sometimes you need a more granular level. Sometimes you need a less granular level. So you can choose from a variety of methods to do that. You can you can, try to a variety of methods for filling data, for upsampling, and a variety of aggregation methods, for for downsampling data. K? On missing data, we've added a a a data function for working with missing data. You can simply apply it to your your data table, and it will report back to you, some some basic things such as the percentage of missing datas in the columns and across the rows. But it'll also give you some interesting insights on what you can do to your data to help you get more to to to have less missing data in the ultimate table. So, for example, they'll tell you if you take out this column, then that will result in 86%. Only 86% of of your data will be or you'll have 86% of your entire data, and the rest might be missing. So it gives you some some things you can look at for strategies on how to how to manage the data, how to how to work with the rows and columns to optimize what you might have and what was originally fairly sparse data. Okay. One thing we're really excited about with Spotfire data science and Spotfire fourteen five is the addition of what we call add ons. So what are add ons? Add ons are new visualizations and new actions that you'll have access to that you can bring into Spotfire and immediately use. Visualizations are pretty clear. Those are new portrayals, new visual new visuals that you can use immediately. Actions, for those of you that might not be aware, in Spotfire fourteen four, we introduced the concept of actions. And an action is you could you could create an action yourself, and you can what you do is you can use it to script Spotfire to do certain things to make things very convenient for someone else to use and then provide a nice input user interface. And then you can bundle that all up and and deploy that and allow others, to use as well. So as an application developer, it's a very, very helpful thing. Now what we're doing is we're providing some of these add ons ourselves that are tailored towards some specific industry problem. And the way that you're gonna be able to get access to these visualizations and action add ons is directly from within the product. There's a new tab. You can click on add ons. It'll immediately show you what's available. You can browse the catalog. You can see a combination of Spotfire data science exclusive add ons, which are available for Spotfire data science users in addition to, add ons that are available in the community as well too, which which we'll continue to provide as well. The Spotfire data science add ons will be covered by the Spotfire warranty. They're considered effectively part of the product, so part of warranty and support. So take a look at some of these as well too. So there'll be initially, there'll be eight visualizations and four new actions, but that we're expecting that number to rapidly increase because we can continuously provide these, as we move along. They do not have to it won't be a new version of Spotfire that needs to happen before you can get access to new visualizations and new actions. So we're gonna be continuously adding new visualizations and new actions, and we're gonna be focused, again, primarily on high-tech manufacturing, and some other cross industry problems as well too. So that's an area of focus for us. So let's talk about specifically what some of these are. So those of you that you've used Spotfire to produce wafer maps, in high-tech manufacturing, you can absolutely do this in Spotfire. It can be a quite an involved process, in order to actually configure the map visualizations to portray what you need for the wafers. You you have to go through the map configuration. You have to set everything up very carefully. The action gives you a nice way where you can just supply what's in the data. You know, what are my locations for the what are my locations on the wafer? What are the values that I wanna portray? Very simple user interface to get started. It will do all the complex configuration for you, and it will result in a in a in a in a visualization that is the wafer map. And then you can go in and see what it did, and you can modify from there to tailor it to what to tailor the representation to exactly what you need. So actions are are enormous time savers, and and we'll be providing some of these so you don't even have to build the action yourself. You can just use it. Something that's a little bit more sophisticated, is our generate wafer zone analysis. So in this action, very commonly, you will wanna find out where are the patterns of defects on the wafers. And there's different kinds of patterns that can happen. It could be radio or or or angular places where we want to see where these defects are occurring. So the action allows you to specify what the configuration of either radial or an angular angular pattern. You can tell tell the action, you know, how you want to set that up. And then, it will determine the distribution of bin categorizations across the patterns within with across the zones that are defined in whatever pattern that you've chosen. And it will create the visualization that you see on the right there where it gives you an interactive linked visualization where you can see the profile, the bin distributions, and that's automatically linked to a display of the wafers below. So you can see, for example, if you've got a pattern along the green there, we got an interesting thing happening in the beginning, Where exactly would those be located relative to the patterns on the wafers in the visualization below? And that can quickly get you to some insights as to where what to fur further explore to see where the where there might be problems. Right? The violin plot, some of you might have used the the violin plot already. This is a community, add on. It's going to remain as a community add on. It'll be available, of course, through the add on browser if you don't have it to to to to download and use. The violin plot is a nice way to see more granular detail on distributions. We all have data. We don't need to understand what the distributions are for those, whether it's process data or any other data. A box plot is a handy tool for that. It gives you essentials, like where the hinge points are and so forth. But the violin plot adds much more deeper understanding of the of the of the detailed distributions of of your data. So it uses a kernel density estimation to give you a nice smooth representation of exactly how your distribution is changing across your your your values. You can overlay things like summary statistics so you can get your classic minimums, maximums, averages, and so forth. There's support for log scales in in the mod, and it supports marking as well too. K. So that's a good example of a visualization that we provided. We're also providing actions for working with time series data. So these actions some of the actions will make use of the data functions that we've introduced as well too. This is a good example of those. These actions will allow you to normalize time series so you can get them on the same scale so it's easier to compare data across series. There's an imputation feature as well too. So you can there's a variety of techniques that are provided to help you fill in the gaps and impute the data, including interpolation or the previous value, next value, or some summary statistics that could that could help you fill in what the missing values are. And, also, the resampling data function that I I referred to earlier, you can use that through this action as well too. So you can instantly, resample the data, to what you need. And, finally, the smoothing, is is also used in this action as well too. So you can easily see what the smooth values are on top of the line that you're working with. And it'll also, additionally, it'll do some calculations to see the distances between where the outlined values might be and the predicted smooths might be, and that can give you a sense of your distribution of outliers as well in your, in your time series. K? And for missing data, again, it's gonna use the missing data data function. It'll give you a nice visual portrayal of the of what the missing data data function provides. And so you can easily see, you know, where are the missing data, where the, you know, the rows and columns that I need to deal with, and also the, you know, some of the insights on on which columns or or rows that you might be able to remove to to get to that optimal least sparseness, in your data. K? Now part of Spotfire data science as well, is is Spotfire Statistica. Spotfire Statistica, if you're not aware, this is a a a advanced data science tool. There's decades of development in that has gone on to into Spotfire Statistica, providing very robust algorithms. It's an end to end life cycle data science platform, supports the full process. It's got a low code, no code, interface where you can specify the operations you want and connect the data flow resulting in a visual analytic workflow. Very widely used in manufacturing, especially pharmaceutical manufacturing, but there's a lot of general manufacturing capabilities that are inside of inside of Spotfire Statistica. And it's got the the the full span of data preparation, classical statistics, advanced analytics, industrial statistics, machine learning. And what's important about Spotfire Statistica, is that it has a nice integration with Spotfire itself so you can leverage those routines directly from Spotfire. So you can create a Spotfire data function that uses the capabilities and Statistica along with your for your to to solve your problems. And so with that, I think I'm gonna turn it over to, Tomáš, so he can give us a demo on a lot of these. Thank you. Okay. Cool. So, yes. So as was mentioned, Spotfire is used by global semiconductor companies. So, also, the demo will be from that, area. And I will show, some of the new features in action and describe how they can help in various places in semiconductor manufacturing process. So let me set up a stage first. So on the screen, we have one variant of, the yield analysis that is typically done by by our customers in some shape or form. You are looking at the yield information over time or over lots. In our case, it's over lot. If you do not know what means your your percentage, which is on y axis here on this upper graph, that's simply the the percentage of the products which are okay after the whole manufacturing and whole testing. So we can see when it's coming, like, in time, in fact, here. We can see that these last four lots are kind of lower yield, so we can drill down and then check, the details. So the bottom graph, for example, is is about, like, yield percentage for separate wafers. And then on on the right, we have, for example, what parameters are what test parameters are responsible for that low yield, which means which parameters failed the most. So there are various things you would like to do after this initial check because this is really some initial check. And for example, you can review the distribution of these problematic test parameters. Right? And for that, we will leverage new and very useful visual mod, already mentioned, violin plot. So the this visual model is in fact, as as it was mentioned, like, relative to the to the box plot. And the main advantage is that it has also, the way how to show, the actual distribution. So these are these violins there. And so it's not only boxes. And here in this picture, we have, the distribution per lot, which means we can see some changes of the distribution in time. We can see, for example, that, we have some, say, location change starting, let's say, at lot, L 2467. And then we have some even, like, change of the distribution for these last those follow ups. They are really different distribution. Even we can see that it's, like, multimodal. And if we, only prove that, really these violins are useful, if we switch down that violins, we can we can guess maybe that the distribution is is different, but it's hard for only from these boxes. But for sure, we cannot guess that there is something like multimodal distribution. So violin's already definitely adding some value. And now how you can use this plot yourself? If you did not use it, if you want to use it for the first time, as I mentioned, you will go to plus, add ons, and here you will find it in the palette of of additional visuals you have available. So it's over there. And once you use it, then afterwards, it should be available directly in your palette of your of your classical visualization types here. So if I click on violin plot here, I have directly that plot on my screen. I can define it in the way how I want, and and in fact, it's there. So very easy. Violin plot is, of course, only, like, one plot we are showing, but only this standalone plot can be useful in in many other scenarios. So for example, in the in the high-tech manufacturing, it can show the distribution of critical parameters like line width, edge depth, resist thickness. You can show densities for for lots. You can use it, for analyzing electrical parameters or threshold voltage across multiple, dies or wafer or do a yield breakdown by different machines, like for lithography steppers or etchors. And so in this part of the demo, we have shown, the the visual mode. And in fact, visual modes are very easy to adopt because they are simply available in your, in your palette of of visualizations. So with that, we will move to a bit different use case and method. And one of the typical tasks is also to review the process data. These are recorded during, the manufacturing process. Right? And the most of these are, in fact, in the form of time series. So let us leverage some, time series inbuilt features now. For these sensors, we have typically data which might be a bit noisy or hard to interpret. And thanks to that, it's not easy to find directly some trends or outliers. Therefore, we will leverage, in this example, smoothing function for detecting unusual shifts or spikes on sensor data. And we will, in fact, show three ways, how to apply smoothing with these new inbuilt features. So one is the first one is to the data functions. So if I go to f x fly out, you can see that I have right away here some some data functions which are there automatically, so we do not need to install anything. And here we have time series smoothing. So let us try that one. You need to specify, of course, what you would like to smooth. So so in that case, it was, I believe, this one. Smoothing method, you can choose. We will leave default because this is optional parameter. Smoothing level is also optional. And for time series index, we might say that it's it's date because we have that, you know, information in our data. And then we are specifying, like, what to do with this output from the data function, and we would like to add, this output as a new column. So if we look at the background, we can see that, the data function appeared in our data canvas. So we have four inputs. We used only two of them. Then, the time series smoothing is done, and in the end, we have one column as output, which is automatically added to our dataset. You can see that it's it's additional column here. And, of course, you can afterwards pick that, that variable manually and then plot it on the graph. So that was first way, how to how to call smoothing. The other way, maybe more interesting one, is to use action. And, again, if you if you want to use, like, action from from our library, then you can go to add ons and from actions. Or if you already have some in the library or in your analysis, you can go to this new icon here for actions. And here we have smooth time series. And we will use this smooth smooth time series action, and we will define it. And and that's kind of the same. Right? You need to define what you would like to to to do. So, again, we can we can smooth the same, the same variable. For example, we can insert a smoothing level. We can insert again what is our index column, and then we would like to create some visuals as well. And if I run it, it created a new page. It's not the same as before. It's here. That's a new one. So it created this action. In fact, created the whole new page. Also, under the hood, the data function was created, the same one as we used before, but now we have, in addition, everything plotted automatically in here. And at the bottom, we have information about, like, discrepancies, like, what is the difference between the actual value and the smoothed value. So with that, you can easily see, like, what are the places when we have some where we have some spikes, some problematic situation, right, which like, times where we should look at the process, what happened there. So, that was the second way how you can how you can, trigger a smoothing. Let me clean this bit. And the third way is that actions on its own can be attached to your visual. And for that, we have here that other action here, and we will attach this particular action. Or maybe I will cancel it. I will show again. I was drag and dropping here, and drag and dropping means, that you can simply reattach that action. And here, we are attaching that as a button. And it's, this page and this visual and smoothing level again. And we click done. But, visually, nothing happened, but we can see that in the corner, we have a action button symbol. And, also, if you go to properties, you will see that there is there is defined action here. So you can define this action for for any user of your dashboard in the future to to have option to simply click here and do this action only on demand when somebody need that. Right? So they will click, and they will have they will have this, this smoothed curve there. So where smoothing is useful, so it was already mentioned, but it can be used in tool monitoring and predictive maintenance. It can be used in in, like, smoothing wafer yield for trend and degradation analysis, or, for example, run to run control uses some smoother historical data to compare. So there are various places where you can use this. So we have seen, action mode behavior, but let us look at one more example showing even, additional features of action modes. So for that, I will switch to to different dashboard here. And there we, or there are situations where you would like to do also some spatial analysis on the wafer, See spatially, right, where is the where is the problem? What is the real problem? And here, we can see that the bottom part, we can see the wafer maps, and we can see, like, good products or dyes, which is this gray color that's bin one. And then we can see these, red ones, which is bin two and the problematic ones. So we can we can see, like, spatially where they are, but in fact, that's that's it's nice, but we would like to also identify some kind of patterns or or clusters, which have the which has the same behavior. Right? And that's that's kind of the the the visual above is helping in that, and that was already mentioned by then, this, zone profiles analysis. So we have different zones, and and these sorry. And these, these values in these in these particular zones means what is what is the percentage of, let's say, these problematic ones in that particular zone. And if we select zone one, radial zone one, we can see that we have problems mainly in the middle of the wafer. Right? If we select, I don't know, zone eight, radial zone eight, then we have a problem on the edge of that wafer. And you can see they are kind of they have problems on the edge. And if we if we select, the the second part is about angular zones. So if we select one of these zones, like, for example, angular zone four, they have problems in this section, this triangle, like left upper, part of the wafer. So that's zone profiles. You can see that each wafer has kind of, signature here. And the the most important thing here is that the whole page here was, like, calculated and created purely by action modes. And this action modes are, in fact, here, They are called generate wafer zone analysis, and you are doing it in two steps. So first steps first step is to, like, calculate these zones. Alright. So I will not do that because I have already everything done, but, simply you will select what are the coordinates and then how many zones you want for angular and radial zones. And then if you run it, then in fact, in your original dataset, you will have two new calculated columns, which are in fact this, like, assignment of radial zone and angular zone. So this this, action is in fact adding some, custom expression and adding some new cones, some transformations. And the second step you would do is to analyze zone profiles, and this means that here you are, like, defining kind of the same thing, like x y, and also you are defining, like, where are these radial and angular zones. And and this action is, in fact, creating all this page. And in addition to that, it's also, like, creating new data for for plotting that upper graph we have seen on the screen. So very cool that we can do all of this with actions, and I think there is a great potential also for you to create these actions. And before, like, to to a bit, let's say, move forward and enhance the results with an additional analytics, I have one more bit to show, and that that will that is about adding some statistic data function. So we can use STATISTICA in many ways, but here we aim to create some simple way which will leverage and even simplify the information from these zone profiles. So I created the data function, which is using using, these zone profiles these zone profiles as an input, and then, it trying to to plot them in two dimensional space. So I'm using, principal components to to, simply transform this zone profiles. So one point here is one zone profile, one wafer, and it's trying to to plot it in two dimensions. In addition, I'm doing some clustering. So if I have simply that two dimensional graph, I can select some cluster, and I can see that, really, this this group is only, let's say, these which do not have much, problematic, dyes or products. If I highlight some other areas, I can see they are close to each other here at the bottom, and they have kind of the same problem. Right? They are problematic in the middle. These two maybe yeah. I see they're kind of the same. So very, very easy way how to how to even, like, even more simply analyze and and check the patterns and check which wafers are coming together. And only for for curiosity, there is also one output, which is, outline measure. So you can kind of also highlight which are the wafers which are really behaving completely different, to the to the rest. So that was what I wanted to show. To sum up, we showed, how easy it is to use different inbuilt features, including visual mods, data functions, and action mods, and we demonstrated practical benefits for some of these. So I hope and wish you to find features helpful and easy to adopt. And with that, I will hand over to to Alessandro. Okay. Thank you very much, Tomáš. It was very interesting. So, let me walk you through two customer success stories where Spotfire made the difference. One of the top customers we work with is Hemlock Semiconductor. So Hemlock is, really, an amazing story. So let me start and give you an overview on on Hemlock first. If you use a smartwatch, mobile phone, a tablet, or computer, an IoT device, or even solar panels, then you may already be using products from, Ham Hamlox Semiconductor. And then in United States, Hamlox Semiconductor is the largest polysilicon producer, which provides this material to companies worldwide, who then uses it to produce the silicon substrate, used in high-tech electronics and solar panel manufacturing. But not only it's challenging to reduce high purity of polysilicon, but also the process of making polysilicon uses large amount of electricity. So large that Hemlock consumes more electricity than any other single state in Michigan where they have their production plant using up to 400 megawatts during the full production. And for scale, a grocery store uses around one megawatt, and then auto assembly, plant uses around 10 to 20 megawatts. So what was the situation that Hamlok needed to face? Like any, manufacturing context, the path to profitability was cost and waste production and improving at the same time, efficiency and reliability to the end to end process. Especially in the last years, polysilicon became a commodity, so there's virtually no flexibility in pricing. But at the same time, extreme purity of the product is essential to guaranteeing the longevity of the electronics that rely on it. So Hamlok needed to embark on a multi step journey to digitally transform every aspect of its manufacturing process and invested heavily in analytics and data science tools. So once we had empowered Hamlok with Spotfire, the first step was to focus on lowering the overall cost structure. And this required analyzing data from every step of the manufacturing process in order to better understand and quantify the impacts of temperature, pressure, energy usage, and their reactive process. And then also with their power management, program, Hamburg lowered the energy demand on the grid when the consumption was the highest, especially between 11AM and 7PM. And here the result was in saving of 300,000 US dollar per month. So today, not only Hamlok is able to control costs and quality with a data centric approach, but now they can explore new opportunities and create new business models that the company could take advantage of. So for example, as the market for polysilicon diversified and more applications became available for different levels of quality and purity, Hamlok was able to take advantage of this market segmentation by selling the right product to the right market at the right price. Here, for example, solar panel manufacturing are markets requiring less pure polysilicon. So let's take, a look at another great example, from one of our customer, which is STMicroelectronics. So ST has over 5,000 engineers working across 12 fabs and back end S and P sites. That's a huge organization with a lot of complexity, But they found a way to bring their data and expertise together. So with Spotfire, they've rolled out over 1,500 custom built an analytics application and each designed by engineers to solve manufacturing problems from process optimization to visual performance monitoring. And here what's impressive, these apps aren't just used in one place. They are shared across sites making it easier for engineers to use and improve them. That means less duplication, faster insights, and continuous improvement across the entire company. So now engineers, they can explore and act on data without having to wait for IT or data science teams. They've also standardized the way analysis is done across the company, making it easy to share best practices and speak the same data language across various teams. So this is another great example of engineering led innovation at scale and how Spotfire supports both agility and governance across the enterprise. So before closing, a few, takeaways here. As you've seen, visual data science is something new and exciting in the industry. Spotfire combines best in class visual data exploration with advanced manufacturing analytics to speed up problem solving and increase cycles of learning for engineers. And Spotfire helps you to bring together various technical data to find to find insights, apply industry standard advanced analytics, or develop and drop in your own techniques. And finally, Spotfire is designed to scale to large communities of users, allowing them to share their access to data, best practices, and share also their findings. So if you recognize a few of these technical challenges that we have described and you would like to unleash the expertise and fertility of your engineers, we would love to talk with you about more your specific situation and share our experiences from other customers in the semiconductor industry. And with that, I hand it over to JP. Beautiful. Well, thank you very much. Thank you so much to our speakers, Alessandro, Dan, and Tomasz, for the insightful presentation, and thank you to everyone who joined us today. Now before we get on to our q and a segment, just a few things that we wanted to share. So with regards to our webinars, we have two great series that are being added to on a regular basis. So whether you're looking to find out more about Spotfire or looking to learn about the latest in terms of what's new, don't hesitate to register to the full series. In terms of your on demand access to today's session, a recording of today's webinar will be available soon as previously mentioned, so please do keep an eye on your inbox for the link. Now if you're interested in learning more, please feel free to visit our website at spotfire.com or contact us directly. There are lots of ways to interact with us, whether it is via our socials, through our community. Additionally, our blog site has lots of great content where we share the latest on visual data science, dive into Spotfire data science in more detail. And last but not least, if there are any enhancements that you would like to see or have ideas that you would like to share with us, don't hesitate to visit our ideas portal. Now onto our q and a segment. So we've had a a couple of questions that have come through. Now the first question that's come through is, how does Spotfire data science differ from traditional analytics tools used in high-tech manufacturing? Will this question be more for yourself, Alessandro? Yes. I take that. So that's a great question. I I already let me dive a little bit deeper here inside. So, traditional analytics tools in semiconductor manufacturing tend to be a little bit little bit, let's say, too rigid. They're often siloed, by teams or for data types and usually require a lot of manual effort, custom coding, deep statistical knowledge, and they don't really scale across the organization. You might get some, something working for one team or for one department, but it's hard to you reuse or to share. And here, Spotfire is really different. It's built for, visual data science. So you can explore complex datasets interactively, wave our maps, process parameters, test data, and you just name it. You've seen it, also with the demo that, Tomáš presented us. And it's not just about visuals. You can bring also AI and machine learning models, even reuse your own r or Python code all inside the platform at scale. And then we have also Copilot our GenAI extension. It's like having an AI assistant in the room with you. It suggests visualizations, generate functions, and helps explaining what's going on into your data using natural language. But the engineering, is always in control. So it's human plus AI working together. And because it's an enterprise platform, everything that you build, visualization, models, application can be shared, governed, and scaled. So instead of reinventing the wheel every time, you build, you're building it on a solid foundation. I would say in short that Spotfire isn't just another dashboard tool or BI tool. It's, a full on problem solving engine designed for engineers, scientists, and decision makers in high-tech manufacturing. Beautiful. Thank you very much, Alessandro. Our next question, you've already talked a bit to it already in your answer, Alessandro, but, are you leveraging AI innovation for solving data challenges in manufacturing? I'll hand this over to yourself, Alessandro, and Dan, if you wanted to speak to that. Well, yes. I already in my previous answer, I was telling you about Copilot. So it's really a Gen AI tool, that you query. It can intake your documents as well. They are they stay within the platform, and you can use you can query, for example, documentation. You can query how things are done, throughout the platform, or you can also query the your data when you don't know exactly the relation. You just query the, the relation. You're trying to discover what's happening in your data, and it generates you a visualization that you can embed inside, your, data application. Okay. Yeah. And then just add to that, you have a lot of, some people that I've worked with have also, commented on how they've been useful to generate code describing a problem, getting the Python code, and then using it directly within their analysis. It can be a real time saver, and it can it can kinda be a a teaching thing too where you can see what it wrote and and understand what it was doing and then apply it. Perfect. Thank you very much. And I'll just add to that. As you might have seen in our previous slide where we also have webinars, we do actually have a dedicated Spotfire Copilot webinar on the July 1, where we'll be diving into the topic in more detail. So please don't hesitate to join that webinar to get more information and to see Spotfire Copilot in action as well. Now we have already answered this question, in the chat, but I did just wanna bring it up very quickly, in terms of what version of Spotfire was being shown during Tomáš, demo. So that is Spotfire 14 Five. That is an upcoming release, and, that release will be with you, very, very shortly. So please do keep an eye out on our blogs, on our website, and it with your contacts at Spotfire. Now our next question is, how does Spotfire help identify root causes in complex production issues? Now I believe this could be more for yourself, Tomáš. Yeah. So, yeah, root causes, I would say, like, typical problem. Like, finding root causes typical problem in many industries. So I showed, like, part of it. Right? Partially, you can use even, like, visuals and interactivity and drilling down to find something. That's that's one way. But, of course, it's better to use also some methods for that. So in Spotfire, there are some native tools, which has, basic functionality, which can be used for that. But I would I would, propose to to use some some data functions or even some statistical models if you if you want to be sophisticated and do it really, let's say, in a complex proper way. Yeah. I just wanna emphasize that point. It's a good opportunity to make use of the combination of something like Statistica and Spotfire or if you're more comfortable with Python and Spotfire, but just to get that precision that you might need in determining what are the most important factors when for the underlying causes. You know, having a predictive model help drive that is, can be really beneficial. Fantastic. Well, thank you very much again to all our speakers today. Just looking at the q and a panel, we haven't received any further questions. So with that, I would just like to thank everyone once again for joining us today. And if there are any questions that arise after the session, please do not hesitate to get in touch with us, and we'll make sure we get those answered. Now with that, I'd love to wish everybody a great day, and we will close the webinar. Thank you again. Thank you very much.