Databricks Data + AI Summit 2024 | Andrew Zane

Clips
More from Databricks Data + AI Summit 2024

Andrew Zane

Partnerships Manager

unstructured.io

play_circle_outline Discussion about unstructured data and vector databases

play_circle_outline Partnership with Databricks and other companies

Info

Andrew Zane | unstructured.io

Andrew Zane

Partnerships Manager unstructured.io

>> Good afternoon AI fans and welcome back to Databricks Data + AI Summit here in San Francisco. My name is Savannah Peterson. Delighted to be joined by John Furrier for this very swaggy segment we're about to have. Man, I feel like we've learned so much and everyone's so pumped about good data this week.

>> This is a hats-off session for this one. We are going to talk about unstructured data, my favorite topic, retrieval, augmentation, generation vector databases. All of my favorite topics. .

>> I can tell. I can tell can by your glee just how excited you are.

>> Hats off, baby. Hats off.

>> And poor Andrew. Andrew had no idea... Well, first of all, welcome, Andrew. Welcome to the show.

>> Thank you very much. Good to be here.

>> Poor Andrew had no idea he was going to be on the show today, until we went over and charmed you because-

>> The delightful surprise for today. Happy to be here.

>> So, context for the audience. When I walked on the show floor yesterday, I'm always looking for my swag for the Savvy Swag segment and I saw your hats and all of a sudden I thought, "Wait, did I just walk into some cool hipster clothing store or am I at a nerd conference? Because this is really dope." I love that all the hats, y'all can see so many different styles of Unstructured... We got hot dogs, we got flowers, we got popsicles, jello, we got tigers. I mean, really just loving it. Like you said, if anything ever happens to Unstructured, you can go straight into the branding business.

>> That is our backup.

>> But for real, before we get into the meat of the tech, how did you make sure that the brand, that the company was as cool as it is? Because so many other tech companies just miss it.

>> They sure do. I mean, we just have genius marketers. marketing is a genius. She has a great team underneath her. And I think it fits with the motif of our company. This is all different structures, different colors, and no matter what it is, we can structure it. And that's the kind of point of our company is that we dig all of the messy data and we transform it into a uniform JSON. So, that kind of means we can have a lot of fun with our branding and put out all kinds of stickers and hats out there. So, .

>> Yeah, I mean even the long sleeves, the whole thing is just... Is this the original logo?

>> No, no. There was a different one I think a year ago, but this has been the logo for a while for sure.

>> Has it felt really good to see... Because literally everyone's wearing the hats, has it felt good to see them everywhere?

>> I think our marketing team is having a blast for these last few months. I mean, they're so creative. Every conference and I'm going to conference every few weeks, we have new hats, new stickers. It's been really fun.

>> Yeah, that is great. So, just in case folks aren't as familiar as we are, both with your hats as well as the company, tell us about Unstructured.

>> Certainly. So, Unstructured has been around for not quite two years, and we like to call ourselves the ETL for RAG. And as I'm sure your audience knows, RAG, or retrieval-augmented generation, where you have your model orchestration and a vector database. We actually don't really mess with any of that directly. What we're about is getting people access of their unstructured documents into those vector databases. So, we process 35 different file types. It could be PDFs, Word, PowerPoint, whatever it is. You put it into Unstructured, it's like a wood chipper. We turn it all into JSON, and it comes out chunked and wrapped in metadata. So, you basically pick the vector database of your choice and you're off to the races

>> Smiling at the wood chipper. My whole family is from the Midwest, and so Fargo hits home for us very much. And anytime someone says the word wood chipper, I'm just totally derailed.

>> That should be on one of our hats. We come out for next conference.

>> Yes, and I need one.

>> We'll send you one.

>> Yeah, I definitely need one if that's the case, the wood chipper of data that... Oh my gosh. That's wild. I can imagine you're pretty popular right now then.

>> Yeah, it's pretty nice. We feel like the popular kids on the block, for sure, at all these conferences because dealing with honestly the unsexy part of AI and RAG. And a lot of our users and customers, a lot of Fortune 500 companies are coming to us and they've spun up these new AI and gen AI teams, even if they've been around for a year or so. These really smart individuals that are trying to figure out the difficult problems of orchestration and what vector database and retrieval strategies. And those are really tough questions that people are wrestling with and they don't really want to spend the time wrangling the data. And they want to do the testing on actual proprietary unstructured data of these enterprises. So, that's where we come in as like the easy button. Don't worry about those data pipelines. We're open source too, so you can test with us really easily. Get those data converted into JSON and then to your POCs.

>> A lot of people have been talking about vector databases and vector embeds. Obviously, neural networks is different than the way it was before with keywords and the normal way to interpret it. So, the question people come up with is that everyone's got a vector database now. Every day is a new announcement. Mongo's got one, everyone's got, Milvus has one. What is the relationship between the database and the actual neural network piece of it, the math behind the structured data? Do they rely on each other or is it independent?

>> That's a very good question. That's beyond my pay grade, John. I couldn't tell you the exact relationship between the math and neural network and the vector database. I think when it comes to databases, we just try to play nice with anyone downstream. We're very agnostic, so we have built-in connectors to all the ones that your audience knows, top 10, 15 popular ones, and you can bring in your own. And I think with our customers, we're trying to get a little more opinionated about lining up what vector databases, what embedding models, what retrieval strategies work best with what use case or desired outcome. And we are betting that there's going to be kind of a plethora of different best use cases. If you're in one industry, perhaps a different vector database will be better for your data and use case. And if you're in a different industry or a different line of business, maybe it's a different answer over here. So, we're excited to figure that out.

>> So, which wood chipper works with whatever connectors-

>> Yeah, that's the idea....

>> you would pop them in. Okay, got it.

>> Yeah. Yeah, we're trying to be agnostic downstream.

>> Doesn't matter who's throwing the lumber in there. Geez. I'm going to... Sorry. I'm ruined. I'm absolutely ruined for this segment now.

>> You're off the rails on this one.

>> It's great. We're at Databricks. Tell us a little bit about the partnership. You're the partnership manager. Tell us a little bit about that.

>> Yeah, Databricks has been fantastic. I mean, they invested us. They're part of a series B a couple of months ago. They're one of our key partnerships. Honestly, I'm a partnership manager at Unstructured and I spend most of my time thinking about Databricks and talking to them. So, very simply, we connect with them upstream and downstream, probably more useful the downstream. So, getting unstructured data wherever it is, S3 buckets, blob storage, Google Drives, and getting it into Databricks. It was announced Day two, we are compliant or integrated with their Unity Catalog, UC volumes. And so, we're one of the early partners to align with those. So, they're coming out, they have their models, they have vector search capabilities. So, we think that customers together with us, Databricks will be able to crack these gen AI problems.

>> I love that. What's it like having them both as a partner and as an investor?

>> Oh, we get a lot of attention, to be honest. They do this really well. They do this really well. Databricks has been fantastic. Some of the greatest champions we have are in Databricks and they're stopping by our booth. They're excited to see us. They show us a lot of love.

>> That's great. And open source, you mentioned open source, big part of the community, big emphasis on the show. Why is that so core to Unstructured?

>> Well, developers are our main audience, right? It's a very technical product that we're creating or products that we're creating. We started open source two years ago. We have I think over 10 million downloads of our Python package library so far. I think it's around 1.2 million per month now is what we're up to, downloads. So, that's a growing community that we love and people are constantly coming by the booths here saying that they use us all the time to solve problems.

>> They just said that because they want a hat.

>> Oh, we give hats to anyone.

>> Where are you guys based out of? Is there one place or is it-

>> We're fairly modern. We're remote. I'd say the two biggest clusters would be the Bay Area and then Virginia, DC. We do a lot of federal work. I, myself in Seattle. We've got people in New Zealand and Europe.

>> Love that, .

>> We're a modern country.

>> Yeah.

>> Awesome.

>> That's rad. That's fun. How big is the team?

>> About 55 people now and growing.

>> Wow. You're flexing for that size in a good way.

>> Yeah. Yeah.

>> That's rad. Keeping it lean and tidy. How much have you raised?

>> 65 so far. 65 million.

>> Great. What are the partners you see on the horizon? You've got Databricks.

>> 65 million and 50 people, that is nice. I like the math.

>> Take us through partnership ecosystem for you.

>> Sure. Yeah. MongoDB is a key partner for us as well. I mean we have to play with the cloud providers out there. Microsoft was an early investor, so that's a crucial partnership, AWS, Google, all the usual players. And then, a lot of the key vector stores out there. So, Pinecone, Weaviate, all those.

>> Oh, cool.

>> It's a crazy world out there in the AI ecosystem, so we're trying to play nice with everyone.

>> Is Milvus on the list?

>> Oh, of course, of course.

>> All right, cool.

>> He'll list anyone you want to play nice with. I totally get that. You mentioned that though, and we've talked about it a bit on the show. It seems like with a gen AI revolution, there's a lot more collaboration than we've seen before as everybody tries to navigate how to figure it out. I do think, and confirm or deny this, you're going to have your own opinion obviously, but I'm wondering we're starting to see things start to parse in one direction or another. We're starting to see some really core allegiances, companies going head-to-head, starting to get a little dicier. Are you seeing something similar?

>> Sure, sure. I think people are starting to get a little more opinionated. I think people that initially came into market as saying, "We're just doing Vector database offerings and they're moving a little bit of right to the database or a little bit left." And so, people are trying to offer more full RAG packages and I think it's still early to say who's winning in what area and why. But yes, things are changing rapidly. And even just the conversations we're having with our customers this month is very different than the conversations we were having two months ago and radically different than six months ago. So, everyone's just learning as fast as they can and it's evolving very quickly.

>> What are the biggest differences in those conversations?

>> Well, so I mean, six months ago, not many people knew what RAG was. So, it was a lot of educating our own customers and they're saying, "Hey, we think we need to be using RAG. Can you help us?" And now, everyone knows what RAG is. Everyone's spun up at least a couple of proof of concepts. And we're getting to, with some of our customers, passive POC, getting up into production and now dealing with real economic issues of like, "Okay, we can get it running, we get it working, it's very useful, it's very powerful, but how do we make it so it doesn't break the bank when we let all of our 10,000 users hit our RAG application 10 times a day? That could be a very, very huge bill." So, we're getting a little more into the weeds, a little more sophisticated with where we're at.

>> Awesome.

>> That AI literacy and data literacy where you just hit on there with having to educate the customer, it is nice to hear you feel like things are coming to a level now. Are you seeing people move into not just experiments, but into a much larger environments and at scale?

>> Sure, sure. I mean, some people are more ahead and some people further behind. I mean, people are starting their AI journey do that every day and we're having initial conversations. But yeah, we definitely have people that are standing up multiple production-level applications. So, yeah, people all over the place.

>> Seeing it across the board. That's really exciting.

>> Yeah.

>> Yeah, wow. What an exciting time. It's obviously an exciting time for the company. What about our AI future excites you personally the most?

>> It seems to me as a really core technology that has the power to disrupt a lot of different industries and by disrupting businesses in almost every kind of business unit within a company, from HR to sales, to product development, to consumer goods, to research to... I mean, it just seems to be a fundamental shift of technology for society that has... The sky's the limit for the implications. It's pretty exciting stuff. And honestly, our customers that come to us are the ones that are really showing the power of AI because they come to us with the ideas. Like, "Hey, we want build a RAG application that does X, Y, Z." And I've never heard that before. And so, it's really inspiring to have those conversations and kind of lean in and say, "Great, how would we build this? How do we stand this up?"

>> You get to see the future first.

>> Yeah. Yeah, that's the idea. Very cool.

>> It's so fun. We feel that way on theCUBE here too, as we talk to wonderful, brilliant people like you.

>> I bet.

>> Yeah, it's a masterclass in itself sitting here, sometimes I find myself taking notes like, "Oh, I'm going to have to go back and learn more about that." And it's like, "Right, pay attention to the guest." Hold on.

>> 100%. 100%.

>> Just immediately, you go into student mode and it's a good vibe. What's next for you guys?

>> Well, next week we're launching Unstructured Serverless June 20th, so that's going to be one of our premium offerings. So, it's going to be a lot of the same core offering of our open source, but a little bit more premium versions of our model. So, we're just going to be a little more accurate, a little faster. And then, around the corner this summer is our enterprise platform, so that's going to be a full GUI-

>> Exciting....

>> offering. So, that's going to be huge.

>> Wow. Congratulations.

>> So, we're very excited for that. Yeah, thanks.

>> That's so cool. All right, Andrew, because you've been so great on the show already, I'm sure we'll have you back. What do you hope to be able to say... You talked about how quickly the conversation rate is changing with your customers and in general, I'm sure even at the booth. What do you hope to be able to say this time next year that you can't yet say today?

>> I think I would like to get a more specific answer to your question of what is AI doing for our customers and for the future. I want to see not just these POCs getting up, starting production level, I want to see what the impacts have been across the board. I just think the whole society is going to be more mature in their AI journey as a whole. And I'll tell you where we are as a part of that in a year.

>> I love that. We can't wait for it, Andrew. Thank you for squeezing us in to what's undoubtedly a busy week.

>> Yeah, thank you so much for having me. Appreciate it.

>> All right.

>> Yeah, this has been a pleasure and thank you and your whole marketing team. Shout out to the marketing team, you nailed it. That's how you ended up on the show. So, you all deserve a pat on the back and whatever beverage you would like to consume, at least in my opinion. John, thank you as always-

>> No worries....

>> for making sure we make the magic happen and invite fabulous folks like Andrew on the show. And thank you for tuning in to our fantastic Power packed at two days of coverage here, Databricks AI... Excuse me, we are at Databricks AI Summit. My goodness, it has been a mouthful today. My name's Samantha Peterson. You're watching theCUBE, the leading source for enterprise tech news.

Andrew Zane | unstructured.io

>> This is a hats-off session for this one. We are going to talk about unstructured data, my favorite topic, retrieval, augmentation, generation vector databases. All of my favorite topics. .

>> I can tell. I can tell can by your glee just how excited you are.

>> Hats off, baby. Hats off.

>> And poor Andrew. Andrew had no idea... Well, first of all, welcome, Andrew. Welcome to the show.

>> Thank you very much. Good to be here.

>> Poor Andrew had no idea he was going to be on the show today, until we went over and charmed you because-

>> The delightful surprise for today. Happy to be here.

>> That is our backup.

>> But for real, before we get into the meat of the tech, how did you make sure that the brand, that the company was as cool as it is? Because so many other tech companies just miss it.

>> Yeah, I mean even the long sleeves, the whole thing is just... Is this the original logo?

>> No, no. There was a different one I think a year ago, but this has been the logo for a while for sure.

>> Has it felt really good to see... Because literally everyone's wearing the hats, has it felt good to see them everywhere?

>> Yeah, that is great. So, just in case folks aren't as familiar as we are, both with your hats as well as the company, tell us about Unstructured.

>> Smiling at the wood chipper. My whole family is from the Midwest, and so Fargo hits home for us very much. And anytime someone says the word wood chipper, I'm just totally derailed.

>> That should be on one of our hats. We come out for next conference.

>> Yes, and I need one.

>> We'll send you one.

>> Yeah, I definitely need one if that's the case, the wood chipper of data that... Oh my gosh. That's wild. I can imagine you're pretty popular right now then.

>> So, which wood chipper works with whatever connectors-

>> Yeah, that's the idea....

>> you would pop them in. Okay, got it.

>> Yeah. Yeah, we're trying to be agnostic downstream.

>> Doesn't matter who's throwing the lumber in there. Geez. I'm going to... Sorry. I'm ruined. I'm absolutely ruined for this segment now.

>> You're off the rails on this one.

>> It's great. We're at Databricks. Tell us a little bit about the partnership. You're the partnership manager. Tell us a little bit about that.

>> I love that. What's it like having them both as a partner and as an investor?

>> That's great. And open source, you mentioned open source, big part of the community, big emphasis on the show. Why is that so core to Unstructured?

>> They just said that because they want a hat.

>> Oh, we give hats to anyone.

>> Where are you guys based out of? Is there one place or is it-

>> Love that, .

>> We're a modern country.

>> Yeah.

>> Awesome.

>> That's rad. That's fun. How big is the team?

>> About 55 people now and growing.

>> Wow. You're flexing for that size in a good way.

>> Yeah. Yeah.

>> That's rad. Keeping it lean and tidy. How much have you raised?

>> 65 so far. 65 million.

>> Great. What are the partners you see on the horizon? You've got Databricks.

>> 65 million and 50 people, that is nice. I like the math.

>> Take us through partnership ecosystem for you.

>> Oh, cool.

>> It's a crazy world out there in the AI ecosystem, so we're trying to play nice with everyone.

>> Is Milvus on the list?

>> Oh, of course, of course.

>> All right, cool.

>> What are the biggest differences in those conversations?

>> Awesome.

>> Seeing it across the board. That's really exciting.

>> Yeah.

>> Yeah, wow. What an exciting time. It's obviously an exciting time for the company. What about our AI future excites you personally the most?

>> You get to see the future first.

>> Yeah. Yeah, that's the idea. Very cool.

>> It's so fun. We feel that way on theCUBE here too, as we talk to wonderful, brilliant people like you.

>> I bet.

>> 100%. 100%.

>> Just immediately, you go into student mode and it's a good vibe. What's next for you guys?

>> Exciting....

>> offering. So, that's going to be huge.

>> Wow. Congratulations.

>> So, we're very excited for that. Yeah, thanks.

>> I love that. We can't wait for it, Andrew. Thank you for squeezing us in to what's undoubtedly a busy week.

>> Yeah, thank you so much for having me. Appreciate it.

>> All right.

>> No worries....