Powering Contextualized Intelligence with NucliOS, MathCo’s Databricks-Native Platform

Key Takeaways

Hi everybody, my name is Akash Kosor. I'm the chief product officer at Mathco. Um, I'm presenting here on day three over lunch, so I guess I drew the short straw. Uh, but thank you for coming here and thank you for dialing in online to to listen to this presentation. Um, I represent Mathco, which is an AI and analytics services organization. We work with about 60 or so Fortune 500 customers across the US and Europe. Uh we have been in the space for around eight years. Uh we've grown rapidly in these last 8 years. I've myself have been with the company for about six of those eight years. Uh from the very beginning. Uh today I'm here to talk to you about contextualize intelligence with uh a platform nucleus. So let me just get into our presentation today. So at Mathco we focus on uh three different verticals of services that we offer to our clients. Uh the first one is building data foundations. The second one is enabling business decisions and the third one is leveraging the power of AI. uh within building data foundations. The type of work that we do at Mathco is around building common data models, semantic layers, uh data cataloging, quality, uh data governance, uh migrations and optimizations as well as building out data marketplaces uh which is a great way for uh the organizations, data organizations to like give access uh to data sources which are curated with the right kind of like lineage mapping uh over to data scientists or for business intelligence usages. The second thing that uh Mathco focuses on is enabling business decisions. Here we're talking about advanced analytics uh building out machine learning models uh operationalizing them AI models operationalizing those uh building AI applications for business consumption. Finally, the third uh stream of work that we focus on is leveraging the power of AI. Uh here we're talking about enduser applications. We're talking about business applications which could be business intelligence integrated with AI tools, AI assistance, uh business workflows, mapping those out, giving access to business users as well as uh um creating custom web applications which uh go beyond business intelligence and actually allow you to do a lot of work uh on the front end. So I'll be walking you through uh a demo of our platform that enables these three journeys. Uh the first journey I want to talk about is uh the data journey. I think the screen's uh running a little uh slow. Give me a second here. Here we go. Uh when I talk about the data journey, it starts from ingesting raw data sources. uh harmonizing these data sources uh building out a medallion architecture creating a semantic layer uh the data quality framework and the data governance monitoring that happens uh as well as creating end data products like data marketplaces. Now here uh we have enabled accelerators that we bring in. Uh these accelerators are plugandplay onto a datab bricks tech stack. Uh whether it be a unity catalog implementation or using expansive features that the platform offers us, we can enable this data journey for data organizations for for companies. The second one I want to talk about is the AI journey. Uh machine learning has been mainstream for a while now. uh we want to make sure that we are using our accelerators once again native to data bricks to accelerate this path. Uh here we're talking about uh building out analytic data sets performing EDA hypothesis testing building out features on top of the data foundation features that can be used for business intelligence or to plug into machine learning and AI models. uh the whole process of uh model development includes training, tuning, deployment, uh scaling, monitoring the models is all embedded within our platform in the form of accelerators. Uh the AI development or app development sits on top of this layer. uh so this is a very important layer because uh with a strong data foundation and a strong AI I ML layer that's when you really try to like add value to business users uh through decisioning that can happen at the node versus happening at the core. The last thing I'm going to speak about before we jump into a demo of our platform nucleus is enabling a decision journey for organizations. Waiting for the screen to Oops. Sorry. Yes. So, here I'm going to talk about uh four independent uh pieces of work that we do for our clients and and this really helps our clients make business decisions that drive value. Um I'm going to be talking about business workflows which are nothing but a layout of any kind of business process. Um and the solutions that are tied to each of those steps. Uh the second one is business intelligence powered with AI AI assistance uh automated insights chat bots and the like. Um going to also talk about custom web applications. Now these go beyond dashboards and business intelligence. is not just like statist uh stat static reporting but dynamic tools that helps business users run simulations, make decisions, commit those decisions and and move forward in their business process workflow. Uh these are end-to-end applications. They are hosted within our client environment. Uh we are not a product organization. These are built custommade for each of our clients. And I'm going to show you a demo of this as well. And then finally uh in the world of AI what's happening more and more is that business intelligence is taking a backseat. Uh people want information at their fingertips. We have enabled certain features on our platform that allows business users to build out dashboards, build out reports or do their analyses with uh English language prompts. So I will be walking you through a quick demo of our platform highlighting each of these four independent things that I talked about business starting with business workflows. So let me orient you first to what you're seeing here on your screen. U this is an example of what I would call a decision control tower. Uh it obviously will have the right kind of accesses and controls. So depending on your user uh they will be able to access certain sections and won't be able to access certain modules over here. Right? So what you're seeing over here is a decision control uh watchtowwer. This is specifically catered to say a CPG organization. The person that's looking at this is an executive who is a revenue management executive. As you know that revenue, there are certain drivers for revenue. Those drivers are represented in the left panel. On the top you have overarching strategic initiatives, goals, targets as well as certain insights that are coming in pulling in from the data. Right now when I look at the drivers for revenue, you can see that you have price, promo, assortment, distribution, execution. Now if we double click into pricing, there are certain processes that any organization will have around each of these drivers. The specific ones that we're looking at over here are uh a yearly pricing strategy process which is specific to the category. Uh a quarterly or a mid-year price review um as well as measurement of price uh price changes and and sensitivity analysis towards that. Right now each of these business processes are unique to the organization. Uh there is no oneizefits-all which is why all of what you see over here we customize it and we built it for organizations. So you might have three more processes over here uh within your organization. We can layer that in as well. Uh for the purposes of the demo, let's assume that the company we're talking about has these these three broad processes when it comes to pricing as a driver for revenue. Uh what we're going to do now here from the decision control tower is we're going to double click into the quarterly or the midyear price review process. Now most technologies most products will have some sense of uh applications or business intelligence or like sensitivity analyses that you can drive. Uh what I wanted to talk about first is what is the actual process. So by clicking on the process button you can now open up the decision flow uh that the company has to go through to conduct their midyear price review. Here you can see that the decision flow is layered in terms of a process flow map. Each of these steps have certain actions that need to be committed. Uh each of these steps might have some analysis that needs to happen. Might have different owners assigned to each of these steps of the process. But what you're seeing over here is an overarching view of uh the the the process flow itself where the organization is in the journey of conducting these steps as well as if you click on each of these steps you will also get access to certain tools dashboards applications that will allow you to conduct these processes. So the construct plan step is the one that I'm on currently which is uh isol isolating shopper activation and promotion impact and target segments. There are certain solutions that the organization has access to. These solutions by the way could be PowerBI dashboards. Uh they could be custom web applications. They could be an off-the-shelf product. All of those will get bubbled up because they are pertaining to this specific step of the process. So as you can see on that bottom panel there are a few solution options that the owner of the step has access to. The first one being a price analysis dashboard. This is a simple dashboard, a static dashboard uh where you can look at and analyze price changes. The second one is a price simulator. Now this is not a dashboard. Here you can actually run simulations and and ML models in the back end will run and produce scenario results. Over here uh there's a section on pricing insights. There could also be ad hoc charts and deep dives that you want to do. Uh it's not always governed by certain applications and and products. Uh you could also want to extract some data, do some ad hoc reporting, do some deep dive analysis and then commit a decision to this step. Right? So let me click on the first one which is the price analysis dashboard. All of this by the way is built on a custom front end. So these dashboards are uh within our platform except for every pixel that you see over here can be customized. New process steps can be added. New charts, reports can be added. You can also integrate it with any kind of visualization software like Tableau or PowerBI. So right now what I'm looking at is a front end which is your price analyzer. Uh on the left panel you see certain modules that the user has access to. Right now we're on the skew analysis module. All of these views and tables are being powered by underlying data sets as well as outputs of machine learning models. Right? So you have a few charts, a few graphs over here. Again, these are fully customizable. Uh a bar chart can be a line graph uh and vice versa. You can have data tables over here. Many of our clients who work with us will actually tell us that hey this is what we have been developing in Excel but it's completely disconnected to our data sources. Is there a way for you to like, you know, represent this on your on your tool? And the answer is yes. Like I said, every pixel over here is fully customizable. Right now, this is a a very simple use case. The the manager responsible for this step is simply looking at price insights, analyzing price insights. There's some trend charts over here. Now, let's go to the second use case, which is that of a custom web application. On that same left panel, you can see that there is a module called pricing simulator. So let's go there. Now this is beyond a dashboard. This is not static in nature. Uh this is connected to ML models on the back end which are then connected to features and data sources uh on the on the back end as well. Right? So here what the manager is doing is they are trying to run a scenario of price changes. Um as you can see there's an input column for new price. Let's go ahead and make some changes here. So what you want to do is you want to like drop down, plug in a new price and see the results of that simulation. Sorry, I think I'm too far away from the machine for the clicker to work. Apologize about that. So what happens when you uh put in a new scenario? Uh your ML models are running on the back end and and new outputs are being generated in terms of volume, RSV and NSV. Once you commit this scenario, the scenario gets saved successfully and now you can go ahead and compare your baseline scenario to this new scenario that you have created. This is the new scenario which will be represented in the form of the outputs of the scenario. uh there are several summarized insights that that get gen autogenerated with AI. As you can see that you can select from a list of all the scenarios that have been created by you uh depend on the collaboration and sharing process within the organization. These scenarios can also be shared upward or downward. Uh multiple users can be working on this simultaneously. Uh when you select the new scenario, you see the change in the output and what that represents. And then finally uh the most important thing many of these decisions are not made in isolations. You need to export some of this information out. Plug it in a PowerPoint presentation. Plug it in an Excel sheet. Do some advanced deep dive. You can actually extract anything over here and export it out to uh a PDF or PowerPoint and Excel or CSV to do further analysis. So by comparing these scenarios, what we've enabled is a use case of uh scenario simulation happening by your business user on the front end. Again, no access to ML models which are on the back end. They are sitting under the hood. Uh you are now conducting the next step of the process of analyzing your price and simulating some scenarios which will allow you to compare price changes and their consequent outputs. So now let's move on to some of the cool AI features that we've embedded within the platform. These are again out of box. So say for example we have this dashboard, we have the custom application built. This is all residing within your ecosystem. We can also configure uh LLM that your organization has access to to the back end. So here is a very basic AI feature that of creating insights. It's great that you have these charts and graphs, but at the end of the day, you're trying to put an update in a PowerPoint, right? So, what you can do is you can pull up the AI insights feature. You can add some basic prompts in terms of what you're looking for. So, the prompts are over here, the persona of the user that you're catering these prompts towards, any kind of additional context that you don't see on the screen that you want to add in your summary. Um, as well as any additional instructions, very basic prompts that you can put in there. So let's just build this out. So what I've written out as a prompt is to generate a summary with key insights and recommendations from the graphs that you see on the screen. Uh my persona is that of a sales manager. Uh the context that I want to add is that whatever you're summarizing should be related to ROI with an ROI lens. And some additional instruction is that like you know you want to the question is without hallucinations using deterministic language. Now what happens when you generate these insights is you will get a AI summarized uh output which will talk about the summary of what you're seeing here key insights in bullet points as well as recommended actions. Now you can rate this response so that the machine learns and gets better over a period of time. You can also autoconfigure some of these prompts so that you don't have to enter the prompts every time if you're using this as a in a periodic fashion for an operational report. Um there's also a way for you to verify this response to go straight to source uh using activity buttons that we can put over here. You can actually see where this data is coming from, where the source of the data is, what is the confidence level of that data as well. Right? So this is a very basic AI feature. It comes plugandplay. All it's doing it's taking information which is visual in nature and putting it in descriptive language. obviously can be exported into PowerPoint, copy pasted anywhere directly from the application. Right, let's move on to uh a next feature which is very exciting uh which is that of the chatbot. So AI chatbots are the rage. So nothing new over there but what we've done is we've configured this chatbot to talk about not just the application within which it's residing but also the underlying data set. So as a user, you're not just limited to what you're seeing on your screen, but also have access to the underlying data source. Um, a question that is the top asked questions will come in as an auto prompt. If you want to select one of those questions, you can or you can type your own question. Question over here that we're going to look at is what are the top performing SKUs by revenue and volume last quarter. And what it does, it generates the response. Oops, sorry. And publishes right then and there in text format. Uh like I was saying earlier, you can provide feedback whether you like this response, you did not like this response. If you did not like this response, you can put in your feedback so the machine learns and gets better over time. There's also a source activation button. If the response that a question that you've asked is not actually on the screen, it'll actually take you to the data source where it found the response. This could have been a PowerPoint document that you've added about a past decision that you've made or it could be a SQL query that is hitting a data set to get you that response. So it'll point you to the SQL query so that you can create like a lot of governance around it. Right now more and more we're seeing in organizations that people are stepping away from business intelligence and want information at their fingertips. So what we've also enabled over here is that why do you need this information? You're obviously going to create some kind of report, some kind of data story that you're going to present somewhere, right? So what we've done is uh we've also done away with the need for business intelligence. You can go directly into the prompt and say, you know what, this this chart is great. This was for my own work. I did my scenario simulations. I want to generate a pricing report. Uh with that prompt over there, like you can see those are follow-ups that came in after you asked the first question. You can generate a dashboard from scratch. Again, this is a native feature on the Nucleus platform by clicking uh one such button or providing the prompt. Sorry about this. Uh there we go. So you can just put an English language prompt saying generate a pricing report and it just pulls up a report that it thinks will be useful for you. Uh there's an additional ability to continue writing prompts and say, you know what, I love the data table. Not a big fan of the chart. Let's eliminate the chart and replace it with a different chart. So now you're talking to the agent and literally creating the report as you go along. Obviously once you've got if this is an operational report, you probably need it monthly. Once you've got your perfect report, you save it as your favorite and you always have access to it. So you don't have to do the prompts again and again. You can also do deep dives. Say if you're looking at the information for North America and you want to double click down to state level, you can say, can you deep dive into Illinois? And it'll pull up that report. Like I said, the chatbot over here does not only have access to the application that you're working on, but also its underlying data set. So, you might have some information that you need from the data set uh which is not available on your pricing analyzer, which you can then pull up with just an English language prompt. Right? Um this was my presentation. I know I've run out of time, but if you want to know more, if you want to get a live demo, obviously I was limited here by PowerPoint, so I was just showing you static screens, but this is all live and running on our platform, uh please feel free to reach out and and set up some time to get a live nucleus demo. Uh but I wanted to leave you with this, right? AI is here, AI is present. Uh it is how we use AI, how we harness the power of AI, which is the most important thing. There's a lot of like AI experimentation going on, but really like building powerful accelerators that sit on native technology platforms like data bricks is how we've made sure that what we're doing over here is very scalable. Uh like when I first started this demo, I was talking about various different steps, various different use cases. We have a knowledge repository of most commonly occurring analytics and AI use cases across CPG, across retail, manufacturing, life sciences. So whenever we start working with our clients, we're not building these from scratch. We come in with a 50 to 60% pre-built solution and then we customize the remaining way so that you get speed to value. Uh we have stood up such applications along with simulators and chat bots in as little as 3 months. Uh because of the power of acceleration. Thank you so much for listening to me. I drew the short straw today. I'm doing a day three session over lunch. So I can understand that not a lot of folks would come in but hopefully uh many of you can catch this online and uh uh call Mathco if you want to learn more. Thank you so much.