HSDF THE PODCAST

Future Ready Insights into CBP's Emerging Requirements - Part 1

• Homeland Security & Defense Forum

Share episode

Preparing for the future, CBP is prioritizing emerging requirements such as advanced data analytics, AI-driven decision-making, and resilient infrastructure to address dynamic security challenges.

 This episode focuses on the technological challenges and innovations facing Customs and Border Protection (CBP). Experts discuss how existing technologies can enhance productivity and tackle longstanding issues, emphasizing the critical role of AI and data integration for improved operations. 

  • Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP
  • Jeremy Ocheltree, Director, Innovation Team, Office of the Commissioner, CBP 
  • Ryan Riccucci, Division Chief, Enforcement Technology & Operational Programs, CBP 
  • Ed Hicks, Business Development Manager for Federal and Artificial Intelligence, Dell Technologies
  • Justin Doubleday, Federal News Network (moderator)

 This discussion took place at the HSDF’s Border Security Symposium on December 11th, 2024 

Follow HSDF THE PODCAST and never miss latest insider talk on government technology, innovation, and security. Visit the HSDF YouTube channel to view hours of insightful policy discussion. For more information about the Homeland Security & Defense Forum (HSDF), visit hsdf.org.

• Justin Doubleday, Federal News Network (moderator):

Thanks, very much. Thank you to the Homeland Security Defense Forum for the opportunity to moderate this awesome panel here. I'm Justin Doubleday, reporter with Federal News Network, and, yeah, this is a very timely panel here talking about CBP's emerging requirements. I think we've had a lot of discussion around different technologies today Of course, ai, edge computing, counter, uas, uas. I think these folks here have some insights into a lot of these different topics, and so we'll kick things off here by going to Ryan. What do you see as some of the most pressing technology challenges and opportunities facing CBP today?

• Ryan Riccucci, Division Chief, Enforcement Technology & Operational Programs, CBP :

So two of the most innovative pieces of technology in early human history were the wheel and the box. About 4000 BC, Mesopotamia, around 3000 humans figured out to put those together to have carts, chariots, wage war. And I'll fast forward to 1969. We put a man on the moon. We invented the precursor to the internet. In 1970, the first patent was submitted for rolling luggage. So we put a man on the moon and invented the internet, until we figured out that we should put wheels on a box in modern times, so we would have to carry our luggage. And that's my message. Where I think some of the current challenges are is using existing things in new and better ways to improve productivity.

• Ryan Riccucci, Division Chief, Enforcement Technology & Operational Programs, CBP :

2014 I got asked by the Border Patrol to go analyze a problem on the southern border. We had a robust lay down of technology radars, view sheds, cameras. We had an unacceptable amount of gotaways. In spite of that, what we learned was there was a wash, a deep impression made from monsoons that water makes a route like a canyon people can walk through, and the requirement was waterproof sensors. At the time, waterproof sensors weren't really deployable. 2014, we landed a spaceship on a comet to collect samples.

• Ryan Riccucci, Division Chief, Enforcement Technology & Operational Programs, CBP :

And when I went to the capability providers in industry, science and technology. We need waterproof sensors. That's the requirement for the agents. That's a tough one, they said. My point is even time. Now you can see disruptive technologies like Uber put existing things together Text and GPS. Here I am. I need a ride, and I would say that, although there's been tremendous developments and innovation and we need the things that the InVent team is doing, we have to drive progress, but we also need to make our agents and operators more productive by giving them the tools to use what they have now better. That's why I think the biggest challenge is identifying and keeping focus that there's current technologies now we can deploy and make interoperable. That would increase productivity and have an impact like wheels on a box.

• Justin Doubleday, Federal News Network (moderator):

What are some of those technologies? I think you're detailed to CBP HQ right now working on some things. What do you see as some of those future requirements and where technology can meet them?

• Ryan Riccucci, Division Chief, Enforcement Technology & Operational Programs, CBP :

I'm going to say something provocative. I don't really believe there's any new requirements. I think that, like EIC Cooper said earlier today, we need to be more precise in our requirements and, although our operators understand the problem, I would say the opportunities are we need more operator caveman eggheads like me that could be more precise in the requirements we've had since the 80s about making data interoperable, integrating our cameras and sensors, coming up in an age where Josh and I were on the southern border for programs that had those requirements but didn't quite make it. So we have to be more precise on the underlying facts about why are we having challenges to be interoperable and it's being more precise in our words and our language and how we terminology we use to label and tag data so that that can be usable and interoperable across missions, systems and components.

• Justin Doubleday, Federal News Network (moderator):

Josh, I think that leads pretty well into the work you're doing as deputy chief AI officer. I was looking at DHS's AI use cases that were publicly reported earlier this year. Cbp has 16 distinct AI use cases on that website, many of which are already in operations. So I'd love to know a little bit about what CBP has learned so far from some of these use cases and how that's driving kind of the path ahead for AI at CBP.

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

Sure. So I didn't prepare for today because I was spending all morning finalizing our submission for what's due for the December 16th reporting that's going to go out publicly for all related AI as well. So right now, this is the top of mind for everything that we're doing and one of the challenges I'm seeing and one of the challenges I think we're all seeing is that I could talk to you about a lot of the AI we've built and we've done over the years.

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

We could talk about the autonomous surveillance towers that are on the public inventory. We could talk about automated image detection. These are all computer vision applications that are just looking for items of interest within video and we've trained and, to Ryan's point, significant amounts of annotations required to go into that work. We learned that early on and we're learning now that there are companies and players out there that can help us to both automate that or do that on our behalf, because it's a very cumbersome process. Dod learned that significantly under the Maven program. These are challenges we've all learned.

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

But I think within AI right now, ryan's starting the ball in the right direction, saying we have old problems that we need to tackle with new technologies. But when we're talking about artificial intelligence, right now, everybody in this room is talking about large language models. We're not talking about computer vision. We're not talking about other things. We're talking about generative AI. We're talking about LLMs, and I'm a firm believer now that the challenge is not how to identify or what the requirements are. It's letting your mind open enough to identify them, because there are things that we're finding out right now just through simple pilots, simple conversations, that LLMs can do. That we had not previously considered, and the challenges that we have on the border are the same. We need to establish a better domain awareness portfolio. We need to have better situational awareness of the entirety of the border. We need work done at our ports of entry to help us to identify and move towards better detection and certainty of seizure of fentanyl and other synthetic and hard narcotics.

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

In the Office of Trade, we have just a plethora of challenges, but when it comes to large language models, all of a sudden we have every office at CBP has a new requirement for AI. It's no longer just we have this challenge or this challenge. Every office has a challenge that large language models or generative AI can solve. We need to identify that. We need to identify how it may benefit them, because all of those challenges, once solved, buy time back for the mission, and so I think I kind of just sum it all up we don't know what all of our challenges are, and I think I've said this previously You're all finding them in your own offices at work right now. You're finding issues with administrative documentation, you're finding issues with internal IT requests and chat functionality. It's kind of the Wild West in terms of, like, what is the problem you want to solve? Because there are solutions we didn't know existed.

• Justin Doubleday, Federal News Network (moderator):

And what is the role of kind of the chief AI officer in prioritizing and making sure that one office over here isn't buying something that another office over here has already bought and could have been solved by just?

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

sharing that. So that's actually something we are just starting now, right? So one of the upsides is of AI being so popular is we get to fly the plane at 100 miles an hour while we put the wings on it. And today's the day where we're actually actively partnering across CBP right now to start identifying what those use cases would be. We found a lot when we went back in time and looked at the entirety of our use cases across CBP. We found not that they were duplicative efforts, but duplicative naming conventions. We found areas where people want to do AI development and we're now getting to kind of benefit from all that background work we did when we started saying that you want to do a use case that would automate X, you want to do X. Let's put you both together so you can fund a grander scale mission towards the solution here. But what we're doing right now is working across CBP to start identifying those use cases across everywhere.

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

What we also need to do at the chief officer level in the commissioner's office is not to stunt creativity across all of CBP.

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

It's to highlight those use cases that are enterprise challenges and not bespoke problems where we can really do our best to kind of help implement that at an enterprise level to help those grander solutioning challenges.

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

But I think that one of the things that's really going to happen is we've got to crawl, walk, run out right we are.

• Joshua Powell, Deputy Chief AI Officer, Office of the Commissioner, CBP:

One of the main goals that I have to play in AI is in no way to overstep what's done at the Office of Information Technology or our chief information officer signing by wallet. They own cybersecurity space, they own the architecture, they own the IT support for the mission, and rightfully so, because they are wonderful at it. My job is to make sure the role and the access and the technology we give to solve mission challenges is appropriate and accurate and safe for agents and officers to use in their day-to-day mission, because the last thing I want to do is give them a piece of technology that ineffectively makes a decision that or produces a result or an output that negatively impacts the way they do their job, because none of this has been tried in the court of public opinion and none of this has been tried in the court of public opinion and none of this has been tried in the court of law, and we, cbp, need to be at the forefront of accuracy for law enforcement operations.

• Justin Doubleday, Federal News Network (moderator):

Jeremy, we were talking about how the Invent team is really working to kind of connect with Silicon Valley and other areas of innovation to pull in some of this really quickly evolving technology to help meet some of these old CBP challenges and new ones. Get us up to date. Just on, I think it's about year five or year six for Invent. How are things going and how are you looking to address some of these big technology challenges?

• Jeremy Ocheltree, Director, Innovation Team, Office of the Commissioner, CBP :

Yeah, thank you. So let me start with a little bit of, I guess, uniqueness of our team and how we operate. Similar to Josh in the AI officer role, the CBP and invasion team is located in the office of the commissioner. Principal focus is identifying, adapting and delivering commercially available technologies, and there's a certain kind of flavor to the innovation that we do, and it's not exclusively but primarily. We focus on venture capital-backed startups, and the reason for that is, since the middle of the last century, the majority of innovation that's happening in this country.

• Jeremy Ocheltree, Director, Innovation Team, Office of the Commissioner, CBP :

I'm looking at it right, it's an industry. There are folks like Ryan and Josh and others pockets in government that are doing truly amazing things, but if you really want to know where the center of gravity, where the talent and the money is, it's an industry. And so, in 2015, the Department of Defense actually based on a hypothesis that hey, we and I don't speak for DOD, but we DOD we believe that we can leverage commercially available technologies to impact our mission and thus establish the Defense Innovation Unit with the express purpose of engaging with Silicon Valley and other kind of innovation centers in the United States to build a nerve to these relationships, identify technologies and capabilities that the DoD can bring in, and so our mission is very, very similar to that we were established in 2018. We have really robust partnerships with entities in the intelligence community the DoD and DHS S&T.

• Jeremy Ocheltree, Director, Innovation Team, Office of the Commissioner, CBP :

They all kind of fill a different role. One of them is specific for venture capital-backed startups. Specific to DoD where there's overlap in the Venn diagram of a DoD mission and a CBP mission, we'll try to find the commonalities and opportunities to collaborate with them. And then specific to true startups, folks that maybe haven't taken the VC route yet, opportunities to do non-dilutive small business innovation research. There's Silicon Valley Innovation Program that DHS S&T has. That's really geared towards the true startups and so that really kind of gives us a nice combination of operators, covers a lot of bases in addition to kind of what you've heard from OIT and some of the other components and CBP and how they kind of more formally engage with industry. And then we've got some unique contracting authorities that we'll leverage ourselves just organically. Commercial solutions openings there's some great contracting efficiencies that can come from. Again, the SBIR contracts if anyone in here is a small business highly encourage you guys to look out for the SBIR things you'll see from the likes of AFWERX and others. So yeah, lots of great work. Absolutely love it.

• Jeremy Ocheltree, Director, Innovation Team, Office of the Commissioner, CBP :

The focus areas have been largely unchanged, I would say, over the last five or six years and I would say there's a really nice intersection of where we have gaps and needs and where there's really interesting things happening in industry, and those are autonomous systems. These are obviously big, high-level capability buckets, but data connectivity is one that we've been chasing for many years. Our agents and officers and sensors distributed all over the globe and those things need to be connected, particularly in a world where we're having these conversations about highly advanced analytics and artificial intelligence. We need to be connected, particularly in a world where we're having these conversations about highly advanced analytics and artificial intelligence. We need to be able to move information and data. Ai and advanced analytics is a focus area for us. Data and sensors and what type of innovation is happening in those fields. It's a consistent need for us to understand what's happening in our environments. I know you guys have heard there's been multiple speakers from OIT but IT infrastructures and what type of innovation is happening there. It's been this interesting dichotomy of we need to consolidate things in the cloud, while we're also needing to. We're realizing that we want to run these AI algorithms at the edge, like you need edge computing. So we're exploring both of those things and all the connections and you know the architectures that are required to do all these things. There's a lot of innovation that's needed there.

• Jeremy Ocheltree, Director, Innovation Team, Office of the Commissioner, CBP :

And then, lastly, this concept of human performance and resiliency and needing to take care of a workforce and I believe, mr Hawkins, you guys have a talk on that one coming up. So, yeah, lots of great work there in support of all the operational components and the support entities for CBP. Lots of great work there in support of all the operational components and the support entities for CBP. Lots of great things going on there.

• Jeremy Ocheltree, Director, Innovation Team, Office of the Commissioner, CBP :

If I made it on the challenges piece, I just want to add a little bit of maybe additional and I think this has come out multiple times, but I do want to footstomp it. So, particularly with this audience, some of the challenges we need to be able to do meaningful integrations, some of the challenges we need to be able to do meaningful integrations right. So that means, generally speaking, open data standards, open architectures and non-proprietary things. Like we need to deal with the incredible volume and velocity of information that's coming at our agents and officers shrink the OODA loop so they can make decisions quick. I'm sure I wasn't here this morning, but I'm sure when Mr Flores was here, you know talking about the core mission that our agency has and having to balance or walk the fine line between facilitation and security.

• Jeremy Ocheltree, Director, Innovation Team, Office of the Commissioner, CBP :

This is all predicated on making informed decisions as fast as we possibly can to facilitate the trade and travel but keep the bad things out. In order to do that, we need to do things fast. We need to take the information, the data that we have and make it more meaningful for agents and officers. So, yeah, principal challenge just want to foot stop that Again. I think I've heard several of those points made already, but that's a core element for our team and certainly something I want to broadcast to this audience.

• Justin Doubleday, Federal News Network (moderator):

Sure and Ed, I think these gentlemen have set you up well with talking about some of the challenges, some of the opportunities from industry's perspective. How are you proposing to helping your agency customers address some of these challenges and solve them from that perspective?

• Ed Hicks, Business Development Manager for Federal and Artificial Intelligence, Dell Technologies:

So a lot of those challenges are not just specific to CBP or DHS and a lot of them are actually in industry already today. And, to be fair, I came out of the space program so I was one of those people that helped get those rockets up onto the space and reutilizing technology. And if you look at the technology that was required to go ahead and put the space shuttle in orbit, if you look at the technology that was required to go ahead and put the space shuttle in orbit, the technology in that little tiny compute to be able to provide the ability for that to work was very, very small. Now we've gotten to a point where we've got so much more compute, so much more capability, so many more sensors that what we focus on, or what I specifically focus on I come out of the computer vision world, so a lot of the work that you mentioned you know computer vision technologies, building algorithms and such a lot of those were pre-canned, those were structured that were already put in place. But with the technology and the fact that we really have to be able to use that commercial, off-the-shelf type of capability to look at a lot of different things, we really now focus more on how do we repurpose commercial off-the-technology for both government as well as industry, right and that's really where I kind of come in and then allow them to communicate. So it has to be a platform that allows all of that data to be able to be brought in to meet the requirements of whatever the data source may be. A lot of focus around being able to have multiple types of integration.

• Ed Hicks, Business Development Manager for Federal and Artificial Intelligence, Dell Technologies:

Then, as you mentioned, new technology comes in. We get the large language models. We need to be able to utilize that Really. I look at a large language model. The whole chat GPT thing is something more along the lines of the evolution of the keyboard and mouse. We are now no longer having to type. I hate to go to like a Star Trek where you computer, you know tell me what's going on. But we do that with our phone, right? My kid knows how to use and get more information than I do because she knows how to use that phone and be able to ask those questions. So what we're really seeing in the evolution of AI is that movement toward being able to leverage that technology to provide actionable insight as quickly as possible to the user in the most accurate format that we can and as quickly as possible and as securely as possible.

• Ed Hicks, Business Development Manager for Federal and Artificial Intelligence, Dell Technologies:

So we look at things in the industry. My know, my perspective is how do we protect intellectual property? And the government is more, how do we follow the NDAA you know, 2023, 2024, which says AI has to be, you know, has to be homomorphic, encrypted, it has to be federated and it has to have some type of multi, you know, computational aspect of it. So there's got to be some things that are guidelines that we follow. It's the same kind of guidelines that we follow in industry. Reality is, we want to be able to build a technology that can be repurposed for both organizations and to be able to provide that capability of bringing that information out to the edge. So, if you think about a computer vision technology, a manifest for a shipping manifest, et cetera, looking at x-rays, potentially of a vehicle coming in, all of these are technologies that we utilize in the outside world in manufacturing, as an example, if I can take those and I can bring those into CBP and be able to provide the insight very, very quickly, that, hey, that there's an anomaly in this environment that we need to identify, that we need to pull out, and I need to be able to do that as quickly as possible. I want to be able to do that with what we're doing with technology that we see today and that's really where we move toward and what we've looked at federated AI as a primary example.

• Ed Hicks, Business Development Manager for Federal and Artificial Intelligence, Dell Technologies:

A lot of technologies exist today where they were brought into the cloud, but the problem is, as Ryan was mentioning, there's a lot of new sensors that are being put in. All these sensors are coming in. All these sensors are really out at the edge. There's just really no physical way to bring all that data back to a central location, and a lot of it's not necessary. I mean, video is a primary example. If I'm running video and I'm looking at a desert field and there's nothing going on, I looking at a desert field and there's nothing going on, I've just sent a terabyte or four terabytes of data back across the network that has absolutely nothing in it except for maybe a cactus running or rolling through, some type of animal or something like that that we don't really care about. I want to know and I want to have that data that's actionable, that's appropriate, and so looking at federated II, looking at the ability to move all of that data out to different locations, is really what we focus on from the industry perspective.

• Justin Doubleday, Federal News Network (moderator):

All right and Ryan, turning back to you, I heard a lot today. Ai is really all about good data. You can't do AI well, or even at all in some cases, without good data. I know that's something that you're looking at good data I know that's something that you're looking at. How can CBP really take advantage of this increasing amount of data really that it's getting, and clean it up. Do what it needs to do to take advantage of it and going forward.

• Ryan Riccucci, Division Chief, Enforcement Technology & Operational Programs, CBP :

That's a great point because you know big data, I think, is all it really is about. Good data and AI isn't new. I mean 1959 McCarthy, they focused on knowledge representation for expert systems. They focused on being precise and unambiguous in our language and code to use AI, and I think the same thing holds true with our data. But data is meaningless Data in context.

• Ryan Riccucci, Division Chief, Enforcement Technology & Operational Programs, CBP :

You get information and you hear things like we want to connect the dots. I don't know that. I'm interested in connecting the dots. I want to know what that means. So once you add meaning, then you get to knowledge and you want to utilize that knowledge. And I just walked through the steps of knowledge engineering. You have to pull knowledge out of a subject matter expert's head, represent it in words or code in a way that can be useful either for decision making or computing, and I think that's the opportunity and the projects in my piece of this universe which I mean. No one joins the Border Patrol because they plan on being a knowledge engineer, but I work bottom up with folks like Jeremy and Josh to show how you can do knowledge engineering to be precise and unambiguous, how you can code our words and make interoperable and recognize patterns and disparate sensors and disparate data formats, to be able to integrate radars for maritime environment or drones in the air, be able to pluck out patterns when they're violated, and do this in near real time, at the edge, and do useful things like run them against a watch list or compare them against other historical data sets. So that's when I said earlier about there's. These things already exist.

• Ryan Riccucci, Division Chief, Enforcement Technology & Operational Programs, CBP :

Knowledge engineering has been around since AI began and now, I think, being more deliberate. I think we're at a new place where maybe in the 80s, systems engineering took hold. But even system engineering started in 1900s with Bell Technologies, when they had to figure out the most optimal way to deploy a telephone network across the country. 1950 was when MIT started the first systems engineering course and then now every major agency and industry has system engineering to manage risk. Now I would say the opportunity for AI is to focus on the fundamental building blocks of AI, which is representing knowledge accurately and unambiguously in your code and in your terminology and metadata, so that when you have an operator that can explain their problem, we know that we are being precise and unambiguous in representing that problem to make it computable, and the answers are in our data. I think it's missing that translation. Like what again? At XD Coopers I'll have interesting things about understanding what the operators mean and making sure that we're being precise in how we apply that knowledge.