🦄 Making great presentations more accessible.
This project enhances multilingual accessibility and discoverability while preserving the original content. Detailed transcriptions and keyframes capture the nuances and technical insights that convey the full value of each session.
Note: A comprehensive list of re:Invent 2025 transcribed articles is available in this Spreadsheet!
Overview
📖 AWS re:Invent 2025 - Presidio Resonate: AI Agents Orchestrating Multimodal Data at Scale (DAT101)
In this video, a Presidio representative demonstrates an AI-powered content management accelerator that transforms how organizations handle massive media archives. The solution addresses sports leagues' challenges in monetizing decades of footage by creating a unified content repository with intelligent tagging and context-aware search. Using Amazon Neptune knowledge graphs, the platform enables complex queries like "show me every left-handed dunk scored in the last five years where the score was tied," reducing search time from hours to seconds. A live demo showcases how AI agents automatically process video content, extract metadata, and generate highlight clips in 34 seconds. The speaker emphasizes applications beyond sports, including healthcare diagnostics and higher education student tracking. The accelerator includes human-in-the-loop model training and managed services, enabling organizations to compress 39-week marketing campaigns into 17 hours while maintaining creative control.
; This article is entirely auto-generated while preserving the original presentation content as much as possible. Please note that there may be typos or inaccuracies.
Main Part
The Content Chaos Challenge: Breaking Down Silos in Sports Media
Thank you everybody. Well, it's exciting to see at least some people stayed around for a Thursday session. I love it. This is like the perfect day and time because it actually allows me to try out some new material, maybe test land some jokes, land some commentary. It's going to be fantastic. So instead of doing slides and boring you with more slides, what I thought I'd do is walk you through a journey of what's happening and why we created these solutions.
Let's take ourselves back, put ourselves as a producer for a sports league, as an executive for a sports league. What is one of the fundamental problems that they have? And we've heard it over and over again. They have tons and tons of footage. How do you actually use all of your footage? How do you make and monetize that as they go forward? How do you coalesce those across MAM systems? MAMs are media asset management platforms. How do you use your S3 archives effectively? And then how do you bring and start to talk about how you monetize that as you go forward?
Like many organizations, they struggle with how to break down those content silos to create monetization opportunities and for their businesses to grow. Nothing was talking to each other. Nothing actually moves together, and at the same time, they can't even see how they use their content effectively. So this creates a big problem. And when we ask that question, what if your content was connected? What if your solutions were connected with each other? What if your content had context in order to make it forward? And how do we break down the barriers of simplicity, intelligence, and create value across our ecosystem?
So let's step back and walk through what that journey would look like. With all that decades of footage and archived assets, one of the biggest challenges and problems is how do you move that all into a cloud archive? How do you move it into a storage mechanism? It's very difficult and it's very expensive, so there's a barrier right there. But more importantly, how do you organize that data from what you're looking at from metadata? Metadata is a fancy word for tagging, and how do you look at that from disparate information sources?
When I was at other companies, we would have hard drives filled with content. They'd be like Pokemon cards, you'd actually distribute and deliver out to everybody else. Nobody knew where anything was. The problem is once you look at that, you can't ask the question of how do I create a play, how do I create content, how do I create excitement with that footage because it's going to take three, four, five hours to find that particular type of content. You can't launch a new fan experience. You heard from the demo the first day, the keynote the first day, how do you tie your fans to your experience, to your content, to your creators? It's a big opportunity and it's a missed opportunity that they can't sell their archives and their footage. It's called content chaos.
But we want to actually start to simplify and transform that as we go forward and what that looks like is building that central content repository, that brain. We actually look at how we take and simplify the fragmented content and the disconnected assets that are looking at the left hand side, create pipelines and agents to understand what is actually that content doing, what agents you need to kick off, and how you need to process that. And then all the way on the right hand side as you're looking at the screen, you're actually getting that content intelligence, that unified knowledge of how your content and how your information is put together. What this provides for you is the ability to unify and execute your content.
A Three-Stage Solution: Ingest, Enrich, and Activate with AI-Powered Knowledge Graphs
So let's walk through these individual stages, and I'll talk through each of them individually. Think about ingest. Whether you're an HR department, whether you're a medical records company, whether you're looking at sports content and footage that I talked about earlier, you need to figure out where your content is. You need to ingest that, and you want to do it without humans because the more humans are involved, the more error, the more challenges you get with your individual types of content. So the ingest allows you to structure your content, but more importantly, as I talked about earlier, build the right pipelines. Are you looking at sports content? Are you looking at flat files? Are you looking at spreadsheets? How do I tie them all together and ingest them so the data has meaning?
The next part is really around the enrichment, and enrichment is a fancy word for how do we tag and coordinate data. And this is an area where you want to start to build your experimentation with your tagging. And part of the problem is which models do you use. As you heard from a lot of the sessions here, we want to simplify that. So we've built a simple platform to allow you to actually bring all of your models in and start to experiment to see which ones produce the right level of content and information that you need. It's not complex. It's built from a console, so you can actually simply start to put models in and start to leverage that to enrich your data. They handle all the simple pieces of that structure.
What you get out of this structure, going back to my analogy from a sports team, think about a dunk. When you see a dunk on film, it's just a dunk. It loses its context. It loses its contextual awareness. What we're able to do is now put the content, the dunk, in context of what actually happened. So the dunk becomes he got dunked on who? What was the defender? What was the time? What was the scoring position? How did all those things come together to create that? And now you've taken your simple vector searches and created a multimodal search to allow you to understand that content as you go forward.
The last part is activate. Think about it this way. I talked about a sports team, but I'm going to draw the analogy to a large brand or retailer. You work with them. It takes 39 weeks to put together an overall marketing campaign as you look at that. It's okay when you want to activate it. So now you take and compress that 39 weeks by being able to find that content, activate that content effectively, while allowing your creative studios and creative authors to still be able to use and leverage that content as it goes forward. Again, what we did is we didn't make things harder. We used AI and the agents to simplify and uplift their workflow and their processes to make it simpler, easier, and effective for them to work.
How this actually happens, if you think about it from a technical thing, remember that dunk that I talked to you about. Now we've built a knowledge graph. Amazon has really great platforms with their Neptune database to start to put context and information for how all the content works together. So that dunk now becomes a web of meaning. The player, as we talked about, the defender, what kind of play, what happened before the play, what was the scoring opportunity that they looked at. You're starting to build a method to understand how to leverage and use your content effectively.
So now you can ask those complex questions in the activation phase of show me every left-handed dunk scored in the last five years where the score was tied. It's not something you can do today in our complex query system, and it's something that people and producers for sports content want to effectively use quickly, rapidly, and easily. This is a tool that allows you to do that in a very simplified manner. What this does is it takes it from three hours, four hours to a matter of seconds.
So think of yourself, how you actually use this. You actually have to understand the request, and I'm going to show you a quick demo hopefully here shortly of how this all comes together in our platform as an accelerator. But you understand the request, you break it down, you build the right agents as we saw. You could actually build a word agent. You could build a transcription agent. It takes that information, breaks it down for you. It leverages the tools across AWS to parse that information, pull the right clips together. So if you're saying I want 20 seconds worth of footage of dunks over the last three weeks, it pulls the right footage, it pulls the right information, and it starts to assemble that quickly and efficiently.
I call it being a technical director in a box. It's not for those big leagues, but if you're in a break or a commentary section, you can actually push a button, get your highlights, get your clips, and start working forward from that. So let's take a break here for a second. Let me show you the power of this because PowerPoints don't convey the real message of what's happening.
Live Demo and Real-World Impact: Presidio's Accelerator Platform in Action
Think about here. This is the platform that we built. It's an accelerator. It's a tool that you can use as you go forward. All of this content could be random content that you've brought into your archive, from skateboarding content to news clipping to footage. And what do you do with this? How do you process this? How do you use this? How do you add context to what's going forward?
So what do you see here when we process this content for video on demand content? You're going to see a whole set of information that you're going to naturally look at. You're going to look at the tags and the information on the left-hand side. Those are great for simply doing standard searches. Now if you want to take it to the next level, we need to take this and leverage and build context. So now using those tags and that level of information, you can see we're starting to put time frames to the context, to the information. So now you know when a play happened, who the play happened with, what was the context around that, and allows you to start to build complex queries and information so you can start to assemble your content and move very effectively.
Now we didn't do this in a vacuum. We added a human in the loop. So as we heard earlier, we want to keep training our models. If the information isn't correct by the AI and the AI that's generated, you can start to put a human in the loop to start processing that information, and then the model will continually learn as they go forward.
So what do we have here? Remember I talked about earlier that you want to do a simple search, right? So you can actually say fumbles. What this tells you is that in my limited library, you have all the fumbles over the last couple of weeks or months or whatever is stored in your petabytes or terabytes of content. This helps you a little bit. It isolates the content, but as a user, you still have to go back and start to figure out where specific things happen in that context of that frame. This could take you hours. It could take you days. It could take you months.
I remember the first question I asked: can I put together a 22-second clip of all the dunks? This doesn't allow you to do that. But with the power of putting the agents together and the searches together, it allows you to start to build together specific moments in time. So I can't type and talk at the same time. Thank you. So what this is actually doing, remember I talked about the agents, the tools, and the power. We're actually going back and looking at that clip library. We're building the understanding of the architecture to understand what the query is. We're pulling out the specific terms, so Northwestern or fumble, and identifying what you want to see. We want to see a moment in time of when something happened. And immediately, in 34 seconds, you see the power of what took three or four hours naturally displayed, leveraging AWS, leveraging the scale that you have to instantly give you highlights and clips that allow you to move your business forward.
Think about the power of how that simplicity happened and the agents happened to create and optimize this. This could work for an HR system. This could work for a doctor, for example. Think about a doctor in a medical facility that has tons of MRIs and information that they're processing through. A patient comes in with an abnormal cardiac arrhythmia. They can immediately use and leverage the information, search that, and provide a diagnosis to potentially prevent a heart attack. So you can see the power and the ease of use that you have here. And one of the things that we wanted to show by this is that we've lowered the barrier for using AI. We've created the simplicity. We've created the opportunity to start using it and playing with it. So it's not a fear, it's a lift to drive your productivity going forward. Hopefully that makes sense for everybody that's looking at this.
Let's go back to our content here where we're looking at this going forward. We've talked a little bit about the explosion in information here. If you think about how we put this into action and the monetization as we go forward, real-world impacts, what we can see now is that once you can do this, you can start to build sponsorship opportunities. I can go back to a sponsor if you're an NFL team or if you're a basketball team, and you can basically say, hey, show me where Michael Jordan did his dunks displaying his signature red shoes. I want to show that, or you're creating a brand awareness moment where something is happening over a period of time. Now you can actually simply and easily do that.
I was with a major brand before I came to Presidio. What took us 39 weeks to actually put together an entire campaign can take you 17 hours while still maintaining the level of creativity and the autonomy across those content silos to create these great opportunities. And more importantly, you can start to take this and associate this with your CRM systems to create that magic personalization that everybody wants when you talk about marketing and in retail. So I want to talk about how I want to give somebody a specific dunk or a specific layer of content. You can do this now effectively and easily with a simple point and click and do simple publishing out to Facebook and TikTok and other social media platforms, or even put it directly into the subscriber's inbox or put a push notification in.
It also allows you to think about coaches on your staff, whether you're a coach internally or a coach at HR or a coach as an advisor at a higher education institution. You can start to build that pathway to say, how do I understand what a student is doing? Here are all his class grades. Here's all the pieces that he's going to. Here are his test scores. Here are all these disparate pieces of information. I'm going to apply my AI and machine learning model to understand which ones are in trouble, which classes they need to look forward to, and how to help them as they go forward. So it's not only just a productivity tool for sports and agents, but it's a productivity tool across organizations as you go forward.
And I'll start to leave you with what is an accelerator? Because everybody that comes up to me talks about, well, Presidio is developing accelerators. What are you actually doing? We're making it easier, as I talked about. We're lowering the bar. We're creating simplicity. We're creating the ability for you to start to pull in information without having the fear that it goes forward. We take time, right? So we've built the components. We've built the agents. You can start to use that to drive faster time to value. You can build cost optimization, and more importantly, it specifically aligns with what AWS is looking for.
So we've started to introduce a managed service. So if you have the complexity here and you started looking at, well, I don't know how to deploy this, how do I support this as we go forward, we're going to put a managed service in there for you so that as changes happen, as things are moving forward, you can get the new models, you can get the new frameworks, and you can get the new information. So if this is interesting to you, which it is to me, Presidio is a leader as we think about how we work with our systems integrator. We're here to help drive and augment you as you go forward, and hopefully you can see the power and the simplicity we provide to your organization. So if you have more questions, our booth is over in 1210. Come visit us. Come learn more about how we're using accelerators going forward and how we can help your business accelerate itself. Thank you very much.
; This article is entirely auto-generated using Amazon Bedrock.
Top comments (0)