Kirk Marple Podcast Transcript
Kirk Marple joins host Brian Thomas on The Digital Executive Podcast.
Welcome to Coruzant Technologies, Home of The Digital Executive podcast.
Brian Thomas: Welcome to The Digital Executive. Today’s guest is Kirk Marple, former Microsoft General Motors and Stats leader. Kirk Marple is the CEO and founder of Graphlit. Graphlit provides a serverless cloud native platform for automating unstructured data workflows, including data ingestion, knowledge extraction, LLM conversations, semantic search, and alerting.
Graphlit has raised over 4 million in seed funding.
Well, good afternoon, Kurt. Welcome to the show!
Kirk Marple: Yeah, thanks so much. Glad to be here!
Brian Thomas: Absolutely love doing these and Kirk, I appreciate you making the time hailing out of the great state of Washington. They’re in Seattle. I’ve been there, like I said, a few times and just love it.
But again, glad you’re able to jump on and do a podcast with me. So, Kirk, jumping into your first question here, can you share your journey from your roles at Microsoft General Motors to founding Graphlit? What inspired you to start Graphlit?
Kirk Marple: Yeah, I mean, I’ve been, I mean, software developer my whole career and Microsoft was My first job after my master’s degree and kind of worked at a bunch of different startups, bunch of bigger companies, smaller companies, including General Motors, like I said, but been really interested in, I mean, what we call media kind of management at the day at the time.
And now unstructured data is kind of what, what people call it. And, and Graphlit is really all about, you know, how do you extract Information and knowledge from different types of media and unstructured data. So, it’s been a passion of mine for many years.
Brian Thomas: That’s awesome. And I really love that because there’s always a problem that we’re trying to solve and unstructured data over the years.
You and I both share that same space as a developer. And unstructured data, I’m telling you, you know, your, your end users, your business, they need some great reporting or some, some analytics and, and sometimes it’s like, gosh, we got all these data sources and it’s all unstructured. What do we do? So, I appreciate the backstory.
It’s I really do. That really resonates with me. And I know a lot will with our audience as well.
So, Kirk, Graphlit offers a serverless cloud native platform for automating unstructured data workflows. Can you explain how your platform handles data ingestion, knowledge extraction, and LLM conversations?
Kirk Marple: Yeah, for sure. So even taking a step back, I mean, I had a company in the broadcast video space for about 10 years, and I learned there’s sort of this pattern of data ingestion. I mean, you have to get the data from somewhere. Typically, the customer has data in a source somewhere on a storage device, or I mean, a network storage, cloud storage, whatever it is, you know, And start to look at, I mean, it’s not just files, it’s web pages, it’s live streams, audio streams, and having a good kind of funnel of all these different data formats into a system for processing is really key.
And 1 of the things I’ve always looked at is indexing metadata, I mean, making that data findable. By both the metadata that is kind of your author, your title, your keywords, but even anything you can glean from the, the files via AI. And so, we had been doing computer vision work and natural language work to analyze this data.
And started to really continuously kind of make this pipeline for both ingestion and indexing, but then the extraction of knowledge. And that’s where LLMs really fit in where once you have text from a document or text from an audio transcript from audio, you can then prompt the LLMs to give you more information.
It can extract kind of people, places, and things. Or summarize that data, and so we’re using it in various different ways, and it’s kind of an expansion of that data workflow that I’d worked on previously. And now we take that all the way to what we call rag retrieve log minute generation. Which is now taking that extract the data.
Providing it as a prompt with the prompt to LLMs and letting you ask questions about it and gather insights. And so, we now handle that fully end to end process for developers just as an API.
Brian Thomas: That’s awesome. Thank you for sharing. Again, I get really jazzed about people coming up with these amazing ways to make the world a better place.
Leveraging the technology whether it be AI, which is the craze lately, but I do appreciate. You sharing that and break it that down for us and Kirk. How does integrating knowledge graphs with large language, large language models or LLM enhance data processing and retrieval in graph? Let’s platform.
Can you provide an example of practical applications?
Kirk Marple: Yeah, for sure. I mean, it’s, it’s something I had started looking at when I was sort of exploring a podcast discovery platform where I was listening to a lot of podcasts and I was thinking, wow, I mean, there’s information that I’m hearing, but also information of, I mean, the people involved, the companies involved, the topics that are kind of lost, they kind of fall to the ground after you listen to a podcast.
So, I started thinking about, you know, and researching knowledge graphs as a way to sort of define that web of information. And it kind of goes back to the early days of like the semantic web and relationships between these entities. And that’s really what we actually started in kind of building these knowledge graphs with pre LLM technology.
And then it was a perfect fit. Once LLMs came out, it was another way to extract this information. And the value is really in kind of giving color to the questions that you’re, you’re asking that it may be data that’s not literally in the PDF or in the audio transcript, but lives in another data source.
And you have to be able to connect those things up of here’s us talking about, I mean, say, what does Wikipedia say about it? What is your CRM say about Microsoft or another vendor? Or even what is some other data source that we’ve created in our own kind of knowledge management system. And, and that for us is really how you connect all the pieces together in, in these knowledge graphs.
Brian Thomas: Really like that and a lot of times we, you know, we think a particular table has the majority of the data or a particular source has what you need. And at the end of the day, I like how you that analogy of putting some color to it. So, you can really discern when you look at something visually, the human brain, as you know sometimes looks at things a little bit differently.
So, I appreciate that. I really do. And Kirk last question of the day. What are the future plans for Graphlit? How do you envision the company evolving in the next few years, particularly regarding AI and unstructured data workflows?
Kirk Marple: I mean, our goal is really to be an easy-to-use API. I mean, being a developer myself, it’s, I mean, something like Stripe and how good they are at developer experience and their documentation and, and kind of providing an easy path for developers to use them.
That’s really what I’m trying for Graphlit. And so, we want to be that middleware layer where all your unstructured data can arrive. We can create that knowledge representation and then have an easy-to-use API and scalable and performant and all those things to really let developers build even cooler applications that they would have just had to do all this heavy lifting themselves.
And that’s really what we’re seeing from our initial customer set of folks that. Maybe understand what they want and they would have had to build this themselves. And now they found us and they’re I mean, within a day or 2, they’re up and running and now can just focus on their own vertical apps. And that’s really what we love to hear.
Brian Thomas: That’s awesome. Again, you found that solution to a problem, but you’re really making. People’s productivity, the bringing efficiency into their jobs and being able to get really an ROI pretty quickly on leveraging your technology. I appreciate that. And Kirk, it was such a pleasure having you on today. And I look forward to speaking with you real soon.
Kirk Marple: Yeah. Thank you so much. I appreciate the opportunity to be here.
Brian Thomas: Bye for now.
Kirk Marple Podcast Transcript. Listen to the audio on the guest’s podcast page.
