Databricks Data + AI Summit 2024 | Sharad Kumar | Qlik

theCUBE.net

Clips
News
More from Databricks Data + AI Summit 2024

Sharad Kumar

Data Integration Lead for Americas

Qlik

Qlik ignites AI innovation: The power of unified data governance

The democratization of data governance for generative AI is reshaping the landscape of analytics and artificial intelligence.As organizations strive to leverage AI for actionable insights, the emphasis on maintaining high-quality, timely and diverse data becomes paramount, according to Sharad Kumar (pictured), field chief technology officer for data at QlikTech International AB.QlikTech’s Sharad Kumar talks to theCUBE about data governance for generative AI.“What we realized, and I talked to a lot of CIOs and seniors, they realize in order to do AI and gen AI, you need good data,” Kumar said. “It needs to be good quality

play_circle_outline Importance of diverse, timely, accurate, and secure data for AI

play_circle_outline Addressing the Challenges of Fragmented Data Estate: A Unified Solution through Click and Databricks Partnership

play_circle_outline Unlocking AI Value: Nurturing an Effective Data Foundation and Infrastructure for Successful Development

play_circle_outline Use of off-the-shelf models and application development for AI integration

Info

Sharad Kumar | Qlik

Sharad Kumar

Data Integration Lead for Americas Qlik

search

>> Good afternoon nerd fam and welcome back to the Moscone Center. We're here in San Francisco, California at Databricks Data and AI Summit. My name's Savannah Peterson here for The Cube, joined by John Furrier. John, it is bumping in here. Quite literally the music, it's like the Databricks Club.

>> We're on the edge of the network. We're on the front lines of we're getting all the data, sucking into our data lake here at Databricks. It's a great show. Savannah, I mean, I think this is just the beginning of the AI centric world that's coming and this is just the beginning of what is going to change the future of analytics, future of AI evolves. Once the data gets set up, it's going to be fun. Our next guest is going to add a lot of value to that conversation.

>> I am super excited to welcome our next guest, from Click. How you doing?

>> Good. How are you doing Savannah?

>> I am so good. John and I had a great time at your show in Orlando.

>> Yeah, last week. It was quite a party back then last week.

>> It was quite a party. You all had a really good vibe going. You actually had great food at your show, which I do not say very often at these conventions.

>> Did you guys go to Universal Studio? Were you there? No, doing the rides?

>> We had to pop up to AWS Financial Services Symposium. Unfortunately, however, we felt such FOMO. We saw the Dinosaur Day video, super fun. I have a T-Rex dress I really wanted to wear.

>> We went to the customer thing the night before, the pool area, the reception. You guys have great loyal customers.

>> We do. We have the best customers.

>> Your community is amazing.

>> And even our partners, we have the best partners, the best customers.

>> So right now mean the partners want that data analytics. Like I said, we covered your show. You guys have brought agents and the answers are all there. So you guys are on the front end of the changing wave on the analytics side as a platform here. The data is a little bit more down into the data lake, if you will, where the compute all the action is. This is the foundational elements of what will be next generation enterprise scale. So they're working away, selling the picks and shovels to the developers, this open source. What's your take on the show? Give us your read. What's your assessment of the keynote and what's unfolding here?

>> Yeah, we're pretty excited. The keynote was really good. I was pretty excited by it. And a couple of things which we saw coming with the acquisition of Tableau, it's really the convergence of the Delta Lake and the Iceberg format. If you look at the last couple of years, it kind of been divergent formats and even though Databricks has been kind of uniform under the uniform umbrella, they say will make the standards into work, but really the play acquiring tabular. But eventually you bring the two formats into a singular format. So now you have, everybody has an open lakehouse open format. So like Ali described, you can like a USB, you can plug in now different apps and different things onto the same lakehouse. So that will, we saw it coming with the announcement last week. I was pretty judged by the Unity catalog open source. I didn't see that coming. That was pretty cool because that I think starts democratize governance a little bit more. People get a great tool to manage and control the data, their models, and govern them.

>> Well, and everyone talks about democratizing AI. It's not democratizing AI if it's not open source in my opinion.

>> Yeah, yeah, exactly.

>> Otherwise it's a facade of access unless people are actually in the sandbox.

>> Yeah, I like what you said, you don't want to give your data to somebody. You don't want to have them lock in.

>> Yes, he said don't give it to the vendors. He actually used the V word. Don't give it to the vendors and we're a vendor. Don't give it to us either. Like okay, don't see that very often on stage.

>> Yeah. Well , you gave a presentation yesterday on building a trusted data foundation on Databricks. How do you do that? Can you give us a little synopsis of that?

>> Yeah, so what we talked about is look, we are in this age, we are transitioning from years. We have been doing descriptive analytics and BI. Then we moved to AI people were doing training models, but really the gen AI wave over the last 18 months since ChatGPT has really put this on a afterburner. But what we realized, and I talked to a lot of CIOs and senior, they realize in order to do AI and gen AI, you need good data. It needs to be good quality. So what I was presenting yesterday was how you need six principles to ensure what I call goodness in your data and it's fitness for AI. And those principles were around you need to make sure your data is diverse because you want to capture diverse enough data so your analytics doesn't have bias in there. You want to make sure your data is timely up-to-date and accurate because you don't want to build your, again, your chatbot on old data. The customers are interacting with your customer service agent chatbot. You want the freshest, latest data. So that means you need to constantly acquire data in real time and feed it. That's become more relevant. The third principle I talk about is that there needs to be good quality and accurate because your predictions are not going to be accurate if your data is of poor quality. Let me talk about how to need to secure the data. Privacy in the data could be talking about a lot of AI systems work on sensitive information, information that's proprietary to companies, and it's a big danger that getting leaked through the algorithm.

>> I feel like I get a notification every other day for a data leak....

>> risk for companies building that. So that was my fourth principle. The fifth principle was that data needs to be relevant, contextual and findable because if you let's say put data there and nobody can understand what it means, how do I use it in the right context? And then my last principle was it was in a form that can be consumed by gen AI application. That means you need to take the data and apply new things. Like you've got to create embeddings, you've got to vectorize the data put into a vector store like what Databricks GA. So it's consumable by these RAG and LLM driven applications. We're generally talking about how you need these six principles to really create a good foundation and a trusted foundation of data on top of Databricks.

>> Is there any kind of limitations, or I won't say limitate, that may be a bad word, or segmentation around scope of when something gets too big or when do you have to make it bigger? Because sometimes in the medium-sized enterprise, it might be okay to have one solution, but as you grow with your data needs, it might be bigger. How do you think about that as building a foundation? Do you build it too big? Is it too small? I mean, what's the scale? Scope this out for us?

>> Yeah. So what I advise customers when I talk to them is, look, you cannot boil the ocean. You cannot take your entire data and say, okay, I'm going to apply these six principles to my entire data. It's a big problem to solve. So just like us human beings, we break problems into smaller. So my advice is let's say if you look at a domain, and typically when we talk about gen AI, chatbots, people talk about in specific domains. Let's say you apply to customer service. So let's look at the data for the customer service domain and can we apply these six principles to that domain of data. To cleanse it, make it accurate, make it timely so we can start to build more. So we approach it that way, approach it domain by domain, business function by business function as you build it out. So kind of have a roadmap there to break the problem into more manageable-

>> So Ali was on stage, we at the beginning of his keynote. He was laying out kind like, oh yeah, AI is in demand. Obviously the lakes are growing, but he said privacy and security is a big problem. And then the third area was the fragmentation of the data estate as he called it. What is the factors to consider when I'm dealing with this fragmented, basically silos, and old, so you got legacy old and then the classic silos, stove-pipes as Dave calls them. How do you get around that? How do you work around that? Do you just rip and replace? Is it a do over?

>> A couple of things. One, if you look from just purely from a data platform perspective that consolidation has sort of been going on. People, if I look at customers, these have multiple databases in mainframe DB2, Oracle, SQL Server. Now they're going to bring the data into let's say a lakehouse. So that's the first step where you create a unified storage place. You bring all your data together in one place because you can do multiple type of use cases from a singular unified platforms. But I think what I see is still from a data integration management perspective. So he put the picture up with their ecosystem and we are one of the ecosystem partners. So if you look at that, so Databricks has a great, diverse, thriving and growing ecosystem of partners. But that means if I'm a customer, if I'm building an end-to-end platform around Databricks, often you have to pick a different tool to ingest the data, different tool to transform the data, something else to secure the data, something else to do BI. And there's a little plug from Click here. We are one of the few companies who cover and check out multiple boxes in that ecosystem because we are building a unified integrated platform that allows customers to move, transform, and govern the data to the same platform. So that means less point solutions to buy simplified architecture, less complexity in integration and faster time to value.

>> Less headaches and quicker innovation is what I'm hearing.

>> Yeah, exactly. Because it's a big... I mean look, in the past life I was a practitioner. I worked in consulting, helping my customers implement these things. It's a big headache. Taking five of six of these and trying to pack it and put it together, make them work.

>> And they're all different UIs, it's all different-

>> All different UI. So the models don't work together. Five different procurement, they all move at a different pace.

>> So much for user experience.

>> User experience, yeah, exactly.

>> It's interesting. Lake Flow is announced. Obviously that's like what FiveTran does, right? So okay, where in the ecosystem is it safe in your mind and where is Databricks going to take their own territory because at some point they got to start building their own and do they do that in conjunction with the ecosystem? So how's the ecosystem partner? How do you look at where it's safe to plant your flag on the ecosystem and do business?

>> Look, every time as the ecosystem when I look at any of the large vendors like Databricks, they're always going to have capability which look and feel that they overlap or they're competitive to us. But the way I look at it is like Lake Flow. It'll be good for bringing certain kind of data, but if you look at what Click does, that's a bread and butter. We can connect to complex sources like SAP and mainframe, relational data, but we have 200 plus sources SaaS application that we can connect. So we look at Databricks as a company, they're not going to build that. The core bread and butter is building out the Lakehouse platform and the Mosaic AI pieces of it. Yes, they acquired RCN, which formed the basis of a lot of that. But I think as a specialized partner, which does data movement and data transform, we are not worried from that perspective that that's competitive to us. But I can see the point. They need to make it easier for their customers to bring in data, which is what we do for them, which really help our customers hydrate their lakehouse much, much easily and much, much quicker from this diverse set of sources that they have.

>> How long have you been partners with Databricks?

>> I would say multiple years. We have a great thriving partnership. Last year we got the data integration partner of the year award.

>> Well done, love to see it.

>> We have 300 plus partners and growing customer, sorry, and growing. So we are a good, very strong partnership. All our data analytics software portfolio support, a lot of the deep Databricks capability. We work with their engineering staff to make sure we are leveraging a lot of the new capabilities. So our engineering right now is working on leveraging the DBRX, which is their LLM and their vector store. So integrating into all of those to make that available for our customers.

>> That's awesome. Are there any of those 300 customers that you have together that you're allowed to share with us, are there any examples or solutions that you're particularly excited by or proud of?

>> Yeah, so a couple of them. I think one customer was JB Hunt where we worked together. So we actually brought in a lot of mainframe data for them out of the mainframe as well as SQL servers. And a lot of the customer, they're driven by timely access to data. Other example we are very proud of is Mercedes-Benz. So they had this program where they want to improve their overall equipment efficiency, so want to bring data from all their plants together into one place. And they were lacking realtime access to SAP data. And so they selected Databricks' lakehouse. So we are able to bring that SAP data in real time, deliver it so they can start to make real time decisions and build their models on the lakehouse. And they're customers that we are proud of. And we are actually giving out a bike there, which is Trex.

>> Hey yo, we saw the bikes on the floor in Orlando.

>> So we have it one here as well.

>> I want the bike. Got to get over there, scan.

>> So that's another great customer where we pull data from their 400 or 500 stores globally so they can perform, have real time view into the stores and do analytics on the same store analytics. So just a couple of examples of great customers we have jointly.

>> Just a couple of examples, just a couple of brand names people have probably heard about. Yeah, that was great.

>> Yeah, exactly.

>> What's your take on the whole development side of this? Because one of the things we're seeing is a lot more developers want to program with data, not just use data in the app as a database call or whatever. You're starting to see the data quality, generative AI, that's a trend. Is it ready yet? Are apps going to truly have data and code integrating together or are we still in the early days of setting the foundational layer down or what's your vision? Because LLMs are always starting to talk to each other, start to see the data interaction is happening.

>> So just the other thing, we talked about a lot of experimentation going on obviously, which is great, and I love the thing what he said, just get started. And we had that same theme I think last week in Orlando.

>> Jenson said that on stage today.

>> Yeah, exactly. Right. So let's get started. So we see a lot of experimentation, but I think what's going to be very critical to move from experimentation to truly getting value is that's what we talked about, is that you have to have that trusted foundation. So I can build a little experiment in my silo. Data is not secure, data is not good quality. I can try out, stuff is good, but if I have to put it out there, if I have to put that bot as a customer service agent bot, I need to make sure all those principles that talked about it here. So customers are beginning to realize it, that if they don't get the foundation right, it's going to set them back. They're going to do a lot of experiment which look good. They'll say, okay, I need to put into production. Now they're going to turn back and say, oh no, my data is not right. It's not secure, it's not good quality. So some of the customers I talked to now learned from this, and they're beginning to focus on the data foundation while they continue to experiment

>> Everyone evolve. Look at Databricks, they went from Spark and now they keep jumping to that next lily pad because the market's getting bigger and they're just go with the customers and you're seeing a lot more blocking and tackling enterprise stuff, privacy, security, complex states, data states, figure those things out. It feels like we still got a lot more work to do on the infrastructure side with the data engineering.

>> Yeah, and I think what good part, if you look in the last 18 months has changed. So earlier when I talked to customers, everybody was challenged with building their own models. I mean, you need data science experience. You need enough data to train those models. Now a lot of customers we talked to, they're using this thing called RAG, which you don't need. You can use off the shelf model, but really feed your data to the model at the right time to get the answer. So I don't need, so now just turn the focus to, I need application developers who can build these applications, Python based application, and then integrate it with their entire enterprise corpus. So this kind of broad use of AI, a little more mainstream than people who only needed data scientists, which is a specialized area.

>> Well, you guys, your show last week, you guys talk a lot about agents. Here, copilots, the ones where agents, copilot, same thing. That will be basically the data plumber, copilots will be managing these calls, but there seems to be an emphasis. Savannah, remember when we were at two years ago, platform engineering and we introduced the data engineering connection. So cloud native next level is happening with data engineering, but it's highly acute when you think about AI because if you don't lock in and engineer that platform, you're going to be screwed because things will break on top. You can't enable anything.

>> It just won't work.

>> It won't work, yeah.

>> It just doesn't work. All right, this has been fascinating and we're obviously big Click fans. I have a question I'd love to know, we'll probably see you along the way at many shows between now and then, but what do you hope to be able to say the next time we see you at Databricks a year from now that you can't currently say today?

>> That you can, sorry?

>> That you can't say yet today. Where are we going to be in a year?

>> I think you will see a lot more customer success stories around use of GenAI. Because right now, if you look at still our show was there, customer's still talking about plain old analytics. I still go to a lot of shows and we're still lacking customers. So I think one year from now we are going to see a lot more customer stories who have built gen AI applications and truly seeing value from data. I think that's what I expect to see in a year when we're back.

>> Is there a use case that excites you most personally?

>> I think number of them. I feel a couple of areas are very important, which I feel one is supply chain. I feel there's a lot of potential there and obviously around customer service. I feel as I talk to customers, I think that's pretty exciting. And actually some of the simplest one I would put out there, which can be done before that is internally facing. So if you look at product, I talked to product and product manuals, training manuals, because the risk is a little less. Your agents are internal facing as opposed to a customer service bot, which is out there now exposed to your millions of customers you can't afford to take the risk. But taking internal product manuals, we don't have privacy information in there. Converting them into utility tools. I'm a technician on the floor, I want to learn how to do this now. It's a thousand page manual. I can very quickly chat and get the gist of it.

>> That's going to be huge.

>> So I think to me, those applications are probably lowest hanging fruit that people will realize that they're the quickest ones to do and have less risk in them than customer facing.

>> Well, we look forward to discussing the adoption of them at scale with you next time here. John, fabulous time as always. from Click. We'll see you again next time down in... Wait, where's the next Click Connect? Where are we going to, do you know yet?

>> I don't know yet actually.

>> Oh man, I was excited.

>> I'm still recovering from the last one, so I'm not thinking.

>> Well, we're still celebrating what a great time we had in Orlando. More importantly, thank you all for tuning in. We're here in San Francisco at Databricks at Data and AI Summit. My name's Savannah Peterson. You're watching The Cube, the leading source for enterprise tech news.

Sharad Kumar | Qlik

search

>> Good afternoon nerd fam and welcome back to the Moscone Center. We're here in San Francisco, California at Databricks Data and AI Summit. My name's Savannah Peterson here for The Cube, joined by John Furrier. John, it is bumping in here. Quite literally the music, it's like the Databricks Club.

>> We're on the edge of the network. We're on the front lines of we're getting all the data, sucking into our data lake here at Databricks. It's a great show. Savannah, I mean, I think this is just the beginning of the AI centric world that's coming and this is just the beginning of what is going to change the future of analytics, future of AI evolves. Once the data gets set up, it's going to be fun. Our next guest is going to add a lot of value to that conversation.

>> I am super excited to welcome our next guest, from Click. How you doing?

>> Good. How are you doing Savannah?

>> I am so good. John and I had a great time at your show in Orlando.

>> Yeah, last week. It was quite a party back then last week.

>> It was quite a party. You all had a really good vibe going. You actually had great food at your show, which I do not say very often at these conventions.

>> Did you guys go to Universal Studio? Were you there? No, doing the rides?

>> We had to pop up to AWS Financial Services Symposium. Unfortunately, however, we felt such FOMO. We saw the Dinosaur Day video, super fun. I have a T-Rex dress I really wanted to wear.

>> We went to the customer thing the night before, the pool area, the reception. You guys have great loyal customers.

>> We do. We have the best customers.

>> Your community is amazing.

>> And even our partners, we have the best partners, the best customers.

>> So right now mean the partners want that data analytics. Like I said, we covered your show. You guys have brought agents and the answers are all there. So you guys are on the front end of the changing wave on the analytics side as a platform here. The data is a little bit more down into the data lake, if you will, where the compute all the action is. This is the foundational elements of what will be next generation enterprise scale. So they're working away, selling the picks and shovels to the developers, this open source. What's your take on the show? Give us your read. What's your assessment of the keynote and what's unfolding here?

>> Yeah, we're pretty excited. The keynote was really good. I was pretty excited by it. And a couple of things which we saw coming with the acquisition of Tableau, it's really the convergence of the Delta Lake and the Iceberg format. If you look at the last couple of years, it kind of been divergent formats and even though Databricks has been kind of uniform under the uniform umbrella, they say will make the standards into work, but really the play acquiring tabular. But eventually you bring the two formats into a singular format. So now you have, everybody has an open lakehouse open format. So like Ali described, you can like a USB, you can plug in now different apps and different things onto the same lakehouse. So that will, we saw it coming with the announcement last week. I was pretty judged by the Unity catalog open source. I didn't see that coming. That was pretty cool because that I think starts democratize governance a little bit more. People get a great tool to manage and control the data, their models, and govern them.

>> Well, and everyone talks about democratizing AI. It's not democratizing AI if it's not open source in my opinion.

>> Yeah, yeah, exactly.

>> Otherwise it's a facade of access unless people are actually in the sandbox.

>> Yeah, I like what you said, you don't want to give your data to somebody. You don't want to have them lock in.

>> Yes, he said don't give it to the vendors. He actually used the V word. Don't give it to the vendors and we're a vendor. Don't give it to us either. Like okay, don't see that very often on stage.

>> Yeah. Well , you gave a presentation yesterday on building a trusted data foundation on Databricks. How do you do that? Can you give us a little synopsis of that?

>> Yeah, so what we talked about is look, we are in this age, we are transitioning from years. We have been doing descriptive analytics and BI. Then we moved to AI people were doing training models, but really the gen AI wave over the last 18 months since ChatGPT has really put this on a afterburner. But what we realized, and I talked to a lot of CIOs and senior, they realize in order to do AI and gen AI, you need good data. It needs to be good quality. So what I was presenting yesterday was how you need six principles to ensure what I call goodness in your data and it's fitness for AI. And those principles were around you need to make sure your data is diverse because you want to capture diverse enough data so your analytics doesn't have bias in there. You want to make sure your data is timely up-to-date and accurate because you don't want to build your, again, your chatbot on old data. The customers are interacting with your customer service agent chatbot. You want the freshest, latest data. So that means you need to constantly acquire data in real time and feed it. That's become more relevant. The third principle I talk about is that there needs to be good quality and accurate because your predictions are not going to be accurate if your data is of poor quality. Let me talk about how to need to secure the data. Privacy in the data could be talking about a lot of AI systems work on sensitive information, information that's proprietary to companies, and it's a big danger that getting leaked through the algorithm.

>> I feel like I get a notification every other day for a data leak....

>> risk for companies building that. So that was my fourth principle. The fifth principle was that data needs to be relevant, contextual and findable because if you let's say put data there and nobody can understand what it means, how do I use it in the right context? And then my last principle was it was in a form that can be consumed by gen AI application. That means you need to take the data and apply new things. Like you've got to create embeddings, you've got to vectorize the data put into a vector store like what Databricks GA. So it's consumable by these RAG and LLM driven applications. We're generally talking about how you need these six principles to really create a good foundation and a trusted foundation of data on top of Databricks.

>> Is there any kind of limitations, or I won't say limitate, that may be a bad word, or segmentation around scope of when something gets too big or when do you have to make it bigger? Because sometimes in the medium-sized enterprise, it might be okay to have one solution, but as you grow with your data needs, it might be bigger. How do you think about that as building a foundation? Do you build it too big? Is it too small? I mean, what's the scale? Scope this out for us?

>> Yeah. So what I advise customers when I talk to them is, look, you cannot boil the ocean. You cannot take your entire data and say, okay, I'm going to apply these six principles to my entire data. It's a big problem to solve. So just like us human beings, we break problems into smaller. So my advice is let's say if you look at a domain, and typically when we talk about gen AI, chatbots, people talk about in specific domains. Let's say you apply to customer service. So let's look at the data for the customer service domain and can we apply these six principles to that domain of data. To cleanse it, make it accurate, make it timely so we can start to build more. So we approach it that way, approach it domain by domain, business function by business function as you build it out. So kind of have a roadmap there to break the problem into more manageable-

>> So Ali was on stage, we at the beginning of his keynote. He was laying out kind like, oh yeah, AI is in demand. Obviously the lakes are growing, but he said privacy and security is a big problem. And then the third area was the fragmentation of the data estate as he called it. What is the factors to consider when I'm dealing with this fragmented, basically silos, and old, so you got legacy old and then the classic silos, stove-pipes as Dave calls them. How do you get around that? How do you work around that? Do you just rip and replace? Is it a do over?

>> A couple of things. One, if you look from just purely from a data platform perspective that consolidation has sort of been going on. People, if I look at customers, these have multiple databases in mainframe DB2, Oracle, SQL Server. Now they're going to bring the data into let's say a lakehouse. So that's the first step where you create a unified storage place. You bring all your data together in one place because you can do multiple type of use cases from a singular unified platforms. But I think what I see is still from a data integration management perspective. So he put the picture up with their ecosystem and we are one of the ecosystem partners. So if you look at that, so Databricks has a great, diverse, thriving and growing ecosystem of partners. But that means if I'm a customer, if I'm building an end-to-end platform around Databricks, often you have to pick a different tool to ingest the data, different tool to transform the data, something else to secure the data, something else to do BI. And there's a little plug from Click here. We are one of the few companies who cover and check out multiple boxes in that ecosystem because we are building a unified integrated platform that allows customers to move, transform, and govern the data to the same platform. So that means less point solutions to buy simplified architecture, less complexity in integration and faster time to value.

>> Less headaches and quicker innovation is what I'm hearing.

>> Yeah, exactly. Because it's a big... I mean look, in the past life I was a practitioner. I worked in consulting, helping my customers implement these things. It's a big headache. Taking five of six of these and trying to pack it and put it together, make them work.

>> And they're all different UIs, it's all different-

>> All different UI. So the models don't work together. Five different procurement, they all move at a different pace.

>> So much for user experience.

>> User experience, yeah, exactly.

>> It's interesting. Lake Flow is announced. Obviously that's like what FiveTran does, right? So okay, where in the ecosystem is it safe in your mind and where is Databricks going to take their own territory because at some point they got to start building their own and do they do that in conjunction with the ecosystem? So how's the ecosystem partner? How do you look at where it's safe to plant your flag on the ecosystem and do business?

>> Look, every time as the ecosystem when I look at any of the large vendors like Databricks, they're always going to have capability which look and feel that they overlap or they're competitive to us. But the way I look at it is like Lake Flow. It'll be good for bringing certain kind of data, but if you look at what Click does, that's a bread and butter. We can connect to complex sources like SAP and mainframe, relational data, but we have 200 plus sources SaaS application that we can connect. So we look at Databricks as a company, they're not going to build that. The core bread and butter is building out the Lakehouse platform and the Mosaic AI pieces of it. Yes, they acquired RCN, which formed the basis of a lot of that. But I think as a specialized partner, which does data movement and data transform, we are not worried from that perspective that that's competitive to us. But I can see the point. They need to make it easier for their customers to bring in data, which is what we do for them, which really help our customers hydrate their lakehouse much, much easily and much, much quicker from this diverse set of sources that they have.

>> How long have you been partners with Databricks?

>> I would say multiple years. We have a great thriving partnership. Last year we got the data integration partner of the year award.

>> Well done, love to see it.

>> We have 300 plus partners and growing customer, sorry, and growing. So we are a good, very strong partnership. All our data analytics software portfolio support, a lot of the deep Databricks capability. We work with their engineering staff to make sure we are leveraging a lot of the new capabilities. So our engineering right now is working on leveraging the DBRX, which is their LLM and their vector store. So integrating into all of those to make that available for our customers.

>> That's awesome. Are there any of those 300 customers that you have together that you're allowed to share with us, are there any examples or solutions that you're particularly excited by or proud of?

>> Yeah, so a couple of them. I think one customer was JB Hunt where we worked together. So we actually brought in a lot of mainframe data for them out of the mainframe as well as SQL servers. And a lot of the customer, they're driven by timely access to data. Other example we are very proud of is Mercedes-Benz. So they had this program where they want to improve their overall equipment efficiency, so want to bring data from all their plants together into one place. And they were lacking realtime access to SAP data. And so they selected Databricks' lakehouse. So we are able to bring that SAP data in real time, deliver it so they can start to make real time decisions and build their models on the lakehouse. And they're customers that we are proud of. And we are actually giving out a bike there, which is Trex.

>> Hey yo, we saw the bikes on the floor in Orlando.

>> So we have it one here as well.

>> I want the bike. Got to get over there, scan.

>> So that's another great customer where we pull data from their 400 or 500 stores globally so they can perform, have real time view into the stores and do analytics on the same store analytics. So just a couple of examples of great customers we have jointly.

>> Just a couple of examples, just a couple of brand names people have probably heard about. Yeah, that was great.

>> Yeah, exactly.

>> What's your take on the whole development side of this? Because one of the things we're seeing is a lot more developers want to program with data, not just use data in the app as a database call or whatever. You're starting to see the data quality, generative AI, that's a trend. Is it ready yet? Are apps going to truly have data and code integrating together or are we still in the early days of setting the foundational layer down or what's your vision? Because LLMs are always starting to talk to each other, start to see the data interaction is happening.

>> So just the other thing, we talked about a lot of experimentation going on obviously, which is great, and I love the thing what he said, just get started. And we had that same theme I think last week in Orlando.

>> Jenson said that on stage today.

>> Yeah, exactly. Right. So let's get started. So we see a lot of experimentation, but I think what's going to be very critical to move from experimentation to truly getting value is that's what we talked about, is that you have to have that trusted foundation. So I can build a little experiment in my silo. Data is not secure, data is not good quality. I can try out, stuff is good, but if I have to put it out there, if I have to put that bot as a customer service agent bot, I need to make sure all those principles that talked about it here. So customers are beginning to realize it, that if they don't get the foundation right, it's going to set them back. They're going to do a lot of experiment which look good. They'll say, okay, I need to put into production. Now they're going to turn back and say, oh no, my data is not right. It's not secure, it's not good quality. So some of the customers I talked to now learned from this, and they're beginning to focus on the data foundation while they continue to experiment

>> Everyone evolve. Look at Databricks, they went from Spark and now they keep jumping to that next lily pad because the market's getting bigger and they're just go with the customers and you're seeing a lot more blocking and tackling enterprise stuff, privacy, security, complex states, data states, figure those things out. It feels like we still got a lot more work to do on the infrastructure side with the data engineering.

>> Yeah, and I think what good part, if you look in the last 18 months has changed. So earlier when I talked to customers, everybody was challenged with building their own models. I mean, you need data science experience. You need enough data to train those models. Now a lot of customers we talked to, they're using this thing called RAG, which you don't need. You can use off the shelf model, but really feed your data to the model at the right time to get the answer. So I don't need, so now just turn the focus to, I need application developers who can build these applications, Python based application, and then integrate it with their entire enterprise corpus. So this kind of broad use of AI, a little more mainstream than people who only needed data scientists, which is a specialized area.

>> Well, you guys, your show last week, you guys talk a lot about agents. Here, copilots, the ones where agents, copilot, same thing. That will be basically the data plumber, copilots will be managing these calls, but there seems to be an emphasis. Savannah, remember when we were at two years ago, platform engineering and we introduced the data engineering connection. So cloud native next level is happening with data engineering, but it's highly acute when you think about AI because if you don't lock in and engineer that platform, you're going to be screwed because things will break on top. You can't enable anything.

>> It just won't work.

>> It won't work, yeah.

>> It just doesn't work. All right, this has been fascinating and we're obviously big Click fans. I have a question I'd love to know, we'll probably see you along the way at many shows between now and then, but what do you hope to be able to say the next time we see you at Databricks a year from now that you can't currently say today?

>> That you can, sorry?

>> That you can't say yet today. Where are we going to be in a year?

>> I think you will see a lot more customer success stories around use of GenAI. Because right now, if you look at still our show was there, customer's still talking about plain old analytics. I still go to a lot of shows and we're still lacking customers. So I think one year from now we are going to see a lot more customer stories who have built gen AI applications and truly seeing value from data. I think that's what I expect to see in a year when we're back.

>> Is there a use case that excites you most personally?

>> I think number of them. I feel a couple of areas are very important, which I feel one is supply chain. I feel there's a lot of potential there and obviously around customer service. I feel as I talk to customers, I think that's pretty exciting. And actually some of the simplest one I would put out there, which can be done before that is internally facing. So if you look at product, I talked to product and product manuals, training manuals, because the risk is a little less. Your agents are internal facing as opposed to a customer service bot, which is out there now exposed to your millions of customers you can't afford to take the risk. But taking internal product manuals, we don't have privacy information in there. Converting them into utility tools. I'm a technician on the floor, I want to learn how to do this now. It's a thousand page manual. I can very quickly chat and get the gist of it.

>> That's going to be huge.

>> So I think to me, those applications are probably lowest hanging fruit that people will realize that they're the quickest ones to do and have less risk in them than customer facing.

>> Well, we look forward to discussing the adoption of them at scale with you next time here. John, fabulous time as always. from Click. We'll see you again next time down in... Wait, where's the next Click Connect? Where are we going to, do you know yet?

>> I don't know yet actually.

>> Oh man, I was excited.

>> I'm still recovering from the last one, so I'm not thinking.

>> Well, we're still celebrating what a great time we had in Orlando. More importantly, thank you all for tuning in. We're here in San Francisco at Databricks at Data and AI Summit. My name's Savannah Peterson. You're watching The Cube, the leading source for enterprise tech news.