The Deep View: Conversations

Back in 2018, in the wake of the Parkland High School mass shooting,

a group of former Navy SEALs banded together and launched a company

called Zero Eyes. Their mission is

to prevent and mitigate these types of violent

events. And the means that they're going about achieving that mission is

through technology. Here we're talking about artificial intelligence, specifically

object detection algorithms. So they went and they trained

algorithms designed to detect the presence of

guns. Then they combine those algorithms, they

run those algorithms on CCTV cameras. to

flag specific instances of person

with one of those weapons. So my guest

today is Tim Solcer, who is one of ZeroEye's co-founders

and the company's CTO. And we talked beyond just

how the models work and how they were made to work and validated to

work. but into the impacts that they're seeing on

the ground, the impact that they're having, the ways in

which this level of automated security is

changing operations on the ground, and where it all could go.

So, if you are new to the pod, I

am your host, Ian Kreitzberg. If you are not new, you

knew that already, and thanks for being here. This is The

Deep View Conversations. Tim,

thanks so much for joining me today. I appreciate you having me on. Yeah. So,

you know, we've connected before. It's been a

couple of years, which is kind of crazy. But ZeroEyes has been around,

you know, even before that. You guys got your start in 2018. Let's

just start there. And then there's plenty of other stuff to

dive into. But with the kind of foundational

So we got started in 2018 shortly after the shooting in

Parkland, Florida. And that shooting was particularly

terrible because the shooter was in a stairwell underneath

a security camera for, I think, three or five minutes before the

first shot was fired. So we looked at that

and we essentially had this idea of, you know, if

somebody was watching that camera, they would have been able to stop that shooting.

And why aren't we watching cameras? It's basically because there's not enough attention

span and people hours to be able to have eyes on every camera everywhere.

And at the same time, AI was

progressing, object detection was becoming a reality. I'd

worked with some computer vision in a previous startup, and

it was a natural progression to be able to say, you know, if we

can detect any number of objects, faces, dogs,

cats, why can't we detect guns on security cameras and use

that as a means to provide proactive

situational awareness during a shooting. Early days of

ZeroEyes, we got started basically just looking for, can

we build a model to detect guns? That was the first MVP.

And we started off by collecting images from Google

images, web scraping images, open image data set, and

basically trying to build a model that would

detect guns. And we tested on videos like clips

from The Matrix and, you know, random images that we scraped

from the web. We found out really quickly that Those

types of models that are trained off web images

are not easily transferable or generalizable to

security cameras. And so we deployed our first model on a security camera

and the performance was terrible. The images didn't

match up to what the AI was trained on. So the

next step from there was we went out and bought some cheap

security cameras on Amazon and hung them up in our

CEO's backyard, which is where our office was at the time. We were working out

of his basement. And we

found out really quickly that generating our own data in

a realistic environment using a real camera, a

real sensor, was the trick that we needed. And

so we invested most of our early time in the company into

just building a quality model and building an organic data set.

That's the perfect jumping off point, right? I mean, we

hear about this all the time, kind of no matter the application, if you're talking about,

you know, studying whales off the Pacific coast, or in

your case, identifying weapons, it all comes down to data,

the quantity of the data and the quality of the data. So I Tell

me more about that. How much

data did you have to collect? And how did

you think about coming up with different ways of varying

the types of data that you're collecting to making sure

you're getting different angles of, I guess, as great

a variety of weapons as you could kind of conceive of? Yeah,

Being a co-founder with four former Navy SEALs, we

never had a shortage of weapons to use as examples. But

exactly that, garbage in equals garbage out.

And we view our data set as probably the most important thing in the

company right now. In the early days, we never had enough data.

That was 100% of our problem. And

we spent a lot of time traveling to different environments, using

security cameras to collect data and with

different backgrounds, different lighting conditions. And that was

really the challenge that we identified early on was security

cameras are mounted in all different types of environments. Security

camera quality, video quality from cameras varies greatly from

camera to camera and manufacturer to manufacturer. So

it was really important for us to have representative examples of

all of the types of environments that we really wanted to detect guns in.

It started in Mike's backyard and progressed to local schools.

We spent every weekend for months in

the early days traveling to different schools and using their

camera systems to record weapon data. And

then the next step from there was we really

wanted schools to allow us to record data when there was actual

people in the camera views. But

schools were not very willing to have us walk around during

the daytime with guns, with active students in the hallways.

So we then came up with the idea, well, if we can't

collect this data in real time, how can we use the customer

backgrounds to build in some generalization

into the model? And our solution was to build a green

screen AI lab, essentially. So we

built out a 5,000 square foot AI lab with

full green screen walls. And we hung about

100 security cameras throughout the warehouse.

So we were able to walk around with ZeroEyes employees with guns

and then overlay our customer backgrounds behind them to

give that context in the scene that we weren't able

to record in real life. And that's where

we've been for the last few years. But there's a lot of

trends around synthetic data, synthetic data

generation that We'll probably obsolete that at

some point. But we've

invested a lot as a company in having a really high

quality data set that represents as many possible scenarios

So beyond that, you talked a little bit about the algorithms, and

I just want to nail a little deeper into

that. You said back in 2018, we were seeing

advancements in object detection, which is basically, that's

the core thing behind what enables your

technology. That's also one

of the major algorithms in self-driving cars. Um, so

it talked to me about beyond the dataset,

building the system itself, the algorithms, um,

how do you validate that? What was the process

like? And, uh, you know, you're building from scratch. You're

not piggybacking off of other systems. It sounds like,

so I imagine if the data collection process was

intensive, the algorithm construction process.

Yeah, there's a few different steps to it. Because when you think about the

entire video pipeline, the processes of the data, there's many different steps

from simply just decoding the video, then passing

it through an inference engine, and then object tracking. So

it's changed quite a bit over the years. But in the early days, we started with

probably the best object detection technology at the time was fast R-CNN or

faster R-CNN models. We were just starting

to see YOLO models be released, which I think we're on

maybe the eighth or tenth iteration of YOLO models at this point. And

it's kind of diverging quite a bit. But in

the early days, it really affected our hardware processing. I

mean, we were trying to solve this problem of

being able to process real-time video at

the customer's site. Because customers are sensitive to

their security camera data being sent

somewhere else. So we identified really

quickly we have to do this with GPUs. We have to do it with GPUs on

premise. And so the model selection came down to what

model gives us the best balance of accuracy and compute

efficiency. For us, faster RCNN

models were the highest performing at that time. They

were definitely more computationally heavy, which

meant that we couldn't load as many cameras per GPU as we wanted to, which

affected our economics. But over time, the

great thing about research and academia is that they're constantly putting

out new stuff. And for the most part, it's open source. So

we're able to use open source off-the-shelf

models, which has progressed to, I think we're

using Ultralytics models today, YOLOv5

or 8. But with

those advancements in model algorithms, that's

brought with it increases in accuracy. As we're increasing the

quality of our data set, the models, the algorithms themselves

are getting better, and the speed is getting better. So we're able

to run models on more cameras, higher

resolutions, higher frame rates, which all of

those things turn into better detection performance for us. And

we expect that trend to continue in the future. Obviously,

we're still talking about object detection models today, which is kind

of like single frame analysis, but with

the potential for large language models and

vision transformers and things like that, we see a lot

of advancement in the future, not just on the detection side, but also on

the context side. What can we communicate about the

image itself to our customers to provide the best situational intelligence?

Hmm. Yeah. I'm glad you brought that up. Cause I was going to ask you about that. Uh,

but the, the, that big area of advancement that

we've seen over the past couple of years, it's all kind of centered around large

language models, uh, transformer architecture, the vision models

that you, that you mentioned. Um, so that, that

is something that you're exploring early

Yeah, it's something that we see as a huge opportunity because today

we're providing this situational awareness of gun

or no gun. There's either a gun in the image or there's not. And

then we have a human in the loop that allows us to maximize

our detection performance while eliminating false positives. So

customers will never receive a false positive. But

the potential with large language models is now you can extract a little

bit more context from the scene. And you can also provide a

little bit more dynamic inputs. So looking for

more dynamic scenes or different objects in different configurations

that wouldn't be realistic with a single model

We'll stick on that side for a second. You mentioned something else that I want to get into in

a minute, but that'll keep. On

the language model side, the idea of greater context on the images that you're talking

about, right? That's really interesting, because like you said, right

now, gun or no gun, that's kind of it.

Other context, I mean, in a security situation

like that, any added piece of information is

probably going to be very nice to have. But

I wonder, as you start experimenting with

deploying or incorporating additional

models of different types, I

wonder how that changes your validation process. Because

the gun or no gun thing is a little more of a straightforward thing,

it's a different type of model. Language models have

reliability issues and

I wonder how much that might be a challenge or if

there's a way maybe through specific data sets,

small language models, other techniques and safeguards to

Well, I think there's two parts of that. One is Just

like all good security comes in layers, I think the same thing is going to

be applicable for IAI. In

our case, I think there'll be multiple layers of AI models

in the future that process all of the detections that we generate,

but also kind of highlights the value of that human in the loop. If

we were to send a detection directly through to

the end customer and provide some layer of analysis from a

large language model, it's very possible that that's incorrect and

that it could confuse or cause some sort of

issue in the response, the critical incident response. And

today we have a human in the loop that essentially does the same thing, but

they're also capable of performing other actions. So during

a critical scenario, when we see a gun and click dispatch, that's

also initiating a call to 911 dispatch

that's closest to the camera location. And someone on

our team is also getting on the phone with points of contact at the

customer to be able to verbally communicate these things. I

don't see the value of having a human in the loop ever being fully

automated or deprecated because

the value during a critical scenario is we're able to

communicate directly to that POC and not rely on

And so I, I guess what you were just talking about there, that that's the point that

I wanted to get to and make sure we drill down on. Um, cause we've

been kind of talking around it for anyone who's not familiar, um,

with what you guys do, right. It's, it's. Object

detection designed for, uh, you know, to, to scan for

weaponry and, uh, CCTV footage. That's

all linked up with warning systems. You have teams of people

that review flags from the system, but I

would love if you could kind of walk me through, if I was a point of contact, if

I was one of your customers and we put in however

many dozens of cameras we had and

something happens, your camera picks up a

flag, can you just walk me through what

that process looks like from kind of inception of the

model says we might have something here and whatever

I'll take a step back and just talk about it from the entire value chain, because

you're absolutely right. We didn't actually cover that in the beginning. We're

connecting to real-time security cameras, all of the customer's existing security

infrastructure, and we're pulling a RTSP

feed, a real-time streaming protocol feed, from that camera

and running an AI on the video feed frame by frame that's

looking for the presence of a gun. at the point at which our

AI says, it's pretty confident that this object is a gun,

it's going to send that detection to a human in the loop. And the

human in the loop is in our ZOC, our Zero-wise Operation

Center. We have one located just outside of Philadelphia here,

and then another in Honolulu, Hawaii. So we're

able to make use of those time zone differences. The

operators in the ZOC are performing an analysis to

verify whether or not the detection has a real gun in it.

So they see a real gun and they click dispatch. That

dispatch button initiates all of our third-party alerting

methods. So we have a dashboard and

a mobile app ourselves, but we also integrate with

local 911. We integrate with other third-party services

in order to get the detection information to

the customer in the best way that they can utilize it.

So from dispatch, our

ZOC operators will get connected to local 911. They'll

communicate and verify that a local 911 has access

to the alert image, and they're able to generate an

incident based on that. But at the same time, we're also calling

points of contact at the customer site. And what we're trying

to communicate is basically what we're visualizing

on our side. So the

ZOC has a specific script that they stick to, but it's essentially something

like we have an alert, a zeroized weapon

detection alert of what appears to be a

person brandishing a rifle in this setting. That

is really the initiation for an incident

on the customer side. Customers have all different standard operating procedures

of how they want to be notified and what they do following a notification. But

I do foresee more automation in the future around that, where

today we're basically handing off situational awareness to

the customer and allowing them to respond. But

I see a lot of benefit in ZeroEyes being involved

in that response in some way. whether it's something as

simple as providing expertise in

how a school or a commercial office building should respond

to that threat, all the way through to initiating other

forms of response, like dispatching alerts

into access control systems, providing

one-click access control capabilities, so

that customers can actually take action from the ZeroEyes Alert. Um,

Kind of in line with what you were just mentioning about being more involved in

the action. I wonder how far something like this goes,

you know, and, and right. And thinking about where you are right

now, where you might want to go as you keep growing,

since you're analyzing real time footage. Right?

Like that's kind of the whole fundamental crux of

what you're dealing with. If there was an

attack, you know, a person brandishing the rifle, or maybe the person's in

a hallway, they're locked in a room, they're moving. Would

you be able to track the movement of these

kinds of assailants? And in an action stage,

beyond just saying, there's a person here, do something about it,

could you physically or do you physically you

know, speak to local police

officers on the phone and say, okay, this is exactly where the

person is. They're in this hallway. They are doing this right now. They are

moving. They just turned left kind of like an overwatch thing.

Is that, is that a function at all of what you

So yes and no. Uh, we aren't, uh, we

don't have visibility into the customer's camera systems to the extent that

we could actually provide that video or watch a

shooter. walk from camera to camera, and we do that for privacy reasons. Basically,

our AI is the only piece that has access to the live video. But

that being said, you're highlighting on a really key point of our value. And

that is, even after the first shot is fired and we know there's a

gunman on site, we don't know where

he is at any specific point, or the first responders

don't know. And so every time that shooter walks in

front of a new camera that's running zero-wise, and we're able to get a

new detection, we are dispatching that and updating law

enforcement. We do have some local police

departments that use our mobile app and they get the alerts directly to

the mobile app for each individual officer. But

the general feedback we've gotten is that they would prefer to get that

information over the radio from dispatch. And so

our biggest point of contact is that 911 dispatch center. And

our integration with a company called Rapid SOS gives

us the ability to automatically be connected to the

closest 911 dispatch center to where our

customer is located based on their camera location. So that

puts us in direct phone contact with 911 PSAPs,

public safety answering points, and those PSAPs are the ones that

And so because of that integration, any officers on

the ground, as well as points of contact at

whatever building or location you're

dealing with, they know the flag was picked

up by, you know, camera 22B

and hallway seven on the second floor or whatever it is.

So they have a deeper level of situational awareness beyond

Definitely. You also highlight a really important point,

which is a unified mapping interface. Because

if our customer is looking at a map, and we're looking at a different map,

and police responding are looking at a third different map, it

becomes really confusing trying to communicate about landmarks and

where things are located. And so having a unified mapping

interface, which we partner with a company called CRG, Critical

Response Group, and they produce really high

quality interior maps. When

our customer has a CRG map, they're able to provide that to us, and

we are able to overlay that on our dispatch map so that both

the customer and our ZOC operation team is

looking at the same map. And wherever possible, we also include

the local 9-1-1 PSAPs in that when we send our alerts, so

Now, you mentioned for the ZOC, for the operators that

review these flags, that no

false positives, I think is what you said. How are these

operators trained? And I guess, what's

their protocol? in terms of validating that

whatever has been flagged is real or

is not a concern. And I mean, I guess taking

that a step further, if you get a flag that turns out not to be a

gun, but maybe it's something a little weird, is that

I would say the operators back in the ZOC have really difficult jobs.

They are on a daily basis, they see

guns on a daily basis. A lot of times it's

toy guns or ROTC rifles, but they're constantly required

to make split-second decisions, which

is why we focused on hiring veterans and former

law enforcement into the ZOC, because those individuals

back there both have the training and understanding what

a gun looks like and how to respond to a critical scenario, but

also as a part of their former careers as

military and law enforcement, A lot of times they were doing very

similar jobs. They either had watch posts or

they were performing some sort of surveillance where they had to

do basically the same thing and

be able to, during a critical scenario, remain

calm and calmly communicate critical

pieces of information that could be happening at the same time people

are actively under threat. So we've

had a lot of success hiring former law enforcement,

former military personnel into the ZOC. And like

I said, they have a really difficult job. On a daily basis,

they have to understand unique standard operating

procedures from different customers and be able to communicate to the

customer in the best way that the customer needs that information. And

how that translates into real life, I mean, We've

made dozens arrests at this point of people

that had guns in areas that they shouldn't have. But

we also communicate with customers on a daily basis about non-lethal

gun threats. And like I said, every customer has their own SOPs. For

instance, some customers want us to disregard any

toy gun detections. Other customers still want to know about them.

And that extends to a wide array of

scenarios that include all different types of guns

being presented from law enforcement or known trainings,

things like that. So as much as possible, we

try to communicate with our customers so that we're aware of any

reasons that there should be a gun on premise, but generally our customers want

to know regardless of what type of gun it is or what the scenario is.

And that's an awesome point of communication for the ZOC.

So even if we see something that we believe is non-threatening, it's

easy enough just to call up the customer and have that conversation and

Yeah, I'd want to know. I'd want to know. You

mentioned that your work, I

guess, has led to a bunch of arrests of

people who had weapons in places that they shouldn't have had them. I'm

wondering what else you can tell me about the

results that you've seen. Like, we've talked about you've launched

in 2018, you've been in operation, you know, and growing

into new places. what

kind of situations have you come across and

Yeah. So since 2018, we've expanded

to, I think we're up to 47 States. We're spread out throughout

the entire country on K-12, um, public

transit, uh, on the commercial side, we're in, um,

big box retail and logistics centers and things like

that. Um, We

haven't had that stereotypical detection

and arrest of somebody that appeared to be, you know, entering a school to

commit a mass shooting. But we've, you

know, in early days, maybe we would see a gun once a month. We're absolutely

seeing guns, dozens of guns on a daily basis in areas that

I never thought we would see guns. And so the

performance of the model itself has proven itself as

we scale. We're generating detections all the time.

And in the scenarios that we've run with local law enforcement and

law enforcement at customer sites, we've been able to identify that the

reduction in response time is considerable.

So comparing response times

without zeroized alerts, First responders

are basically showing up at a school and not knowing where the

shooter is located. So they enter that school, it

could take them five to 15 minutes to clear the school

and locate the actual shooter. And there's been plenty of real life examples,

like Uvalde, where there's been serious

challenges around that. So in

the testing that we've done, we've been able to considerably reduce response

time and direct first responders exactly where in

the building they should be located. It's

something as simple as if first responders show up to the wrong side

of the building and they enter the wrong door, that could mean 15 minutes

in lost response time. Wherever

possible, we're really focused on reducing that response time to get

You mentioned you're seeing dozens of guns every

day in places that you wouldn't expect. Are

these, what are those kinds

of situations? Is that, are people just kind of caring and

they're walking around and they're caring? Is there intent to

violence? Do those always lead to kind of reactions in

terms of law enforcement, but in terms of, you know, the customer that

you're securing, is there responses to that or is it, oh no,

he's okay. I mean, obviously each

Yeah, short answer is it totally depends. But

I would say that for the most part, fake guns are

starting to look more and more like real guns. If you pick up an airsoft rifle at

your local Walmart, that airsoft rifle is almost indistinguishable

from an actual AR. And even

more so when students will

paint the tips, paint the orange tips black or remove the orange

tips. So a lot of times we

don't necessarily know if the gun is a fake

gun or a real gun, and we have to essentially treat it as though

it's a real gun. So I would say that's probably

one of the more common scenarios that we see. Also, a lot of times we

see people using objects

like cell phones and pointing them at each other as though they're real guns.

There's a very popular TikTok challenge that's been going around the last few

years called Senior Assassin, which is about the

most insensitive thing that I think you could do in nowadays climate. But

students are bringing airsoft rifles or fake pistols,

or in some cases, real guns to school in order to perform

these mock senior assassinations and post them

online. I couldn't tell you how many detections we've gotten that

were similar to that, where we see students basically filming

themselves pointing guns at each other in

order to fulfill this TikTok challenge. So

in cases like that, we always respond to the customer. We let

them know what we're seeing. We try

to communicate as much detail about the scene as possible, but ultimately

it's on the customer to respond and execute

Senior assassins, huh? Yeah. Oh,

man. But and so you mentioned, I

just want to, you know, clarify, right, like you haven't been

involved in or encountered directly

any of the kind of mass violence

Luckily, no, uh, I think about this sometimes and you

know, obviously if an event does happen, I want to be there. I

want to be on those cameras to be able to detect it. Uh,

but thankfully we've, we've not been involved in a mass shooting

event. Um, I think it's only a matter of time though. The

more cameras we're on, the more guns we're going to see, the more coverage we're

going to have. And, um, I just pray that when we're in

that situation, uh, that we're able to make the detection before

Mm hmm. Now,

you mentioned that you're in 47 states now,

and you have a ZOC in Pennsylvania

and in Hawaii. So you cover the

different time zones. They're operational all the time. How

does the ZOC scale in kind with

the scale of your emplacements and your

model? Like how many more people do you need for each

Today, we're able to monitor the entire United States from just those two operating

centers. I anticipate in

the future we're going to expand and build operating centers in other locations, but

today we're able to just basically follow a model that says when we add

X number of cameras, we're going to expect to see X number

of additional alerts, and that increases our headcount. So

we follow a pretty linear model in that sense. We also try

to staff people based on alert load throughout the day. As

you can imagine, we get the most false positives during the times of

day that are most active in front of cameras. So if

you think about a school, that's the five minutes every

hour when students are walking in between classrooms. And so during

the day, probably 8 AM to 5 PM, we

staff heavier than we need to during the nights and off times and

weekends. But it's been

pretty standard for us and the more scale that

we have, the more different sites that we're on, the

more data that we include from those different sites

into our model to make it better, the more predictable

our scaling model is. So that part

Now on the privacy side, which

you kind of referenced earlier when I was talking about the whole Overwatch-y type

thing, have

access to these cameras, but you're not watching those camera feeds

necessarily. You get, you'll get like the frame that

was flagged. But you, you train, like

you just said, the data that you gather, you use to

train the models. And so I, and

I'm sure this probably varies customer to customer, but I'm wondering on

the privacy side, you know, if we're, thinking

about schools or public places, these big box retailers, for

example, how you work on the data side

to ensure privacy and what specific components,

I guess, of these images that they'll

use to train your models on? Is it anonymized in

some way, faces cut or blurred out or whatever?

For our training data, we try not to obfuscate

as much as possible because we're concerned

that it will affect the context of the model. But

for any data that's coming from a customer, we do remove

faces so that we're not infringing on any privacy of

the customer. Obviously, it's something that we get the customer's permission ahead

of time. We're very close with our customers in

that respect. But on the data training

side, we've, from the very beginning, tried to

distance ourselves from any sort of biometric detection

or recognition, specifically to address that privacy concern.

So instead of detecting a person and

then detecting a gun, we're strictly looking for a visible

firearm. And all of our data that is

annotated is annotated for that specific firearm. Avoiding

the facial recognition piece, avoiding any sort of biometric analysis has

allowed us to distance ourselves from privacy concerns with

our customers. And because of that, I think our customers trust us.

They trust that we're not looking at their live video. They

trust that we're not detecting

or biased in any way that would affect

And there's also a security side as well. Obviously, there's

security to everything we're talking about your security company, but the security

of the model, the security of the cameras and the warning system.

And you talked about on prem deployments.

But I'm wondering about, and I'm just gonna throw a

kind of stupid, like movie scenario at

you. Would it be possible

or if If possible, how do you mitigate for

someone to somehow gain access to or

hack into your system to trigger a false alarm? I

can't, I don't know exactly how that would work, but if they were to gain access, like,

is that a real security concern on your end? That

your system could become compromised in a way that sends

I would say that's probably lower on my list of concerns. I

mean, getting hacked in general or falling prey

to some sort of social engineering is always in the back of my

mind and probably is for any technology leader or

startup founder. But

that piece of our business, it always runs through the

ZOC. So the only way for an image to be dispatched to our customers is

for somebody in the ZOC to actually click dispatch

and send it out. Like I said, we have

a really tight connection with our customers. So even in the event that

there was an accidental dispatch, we

would have direct line to the customer to be able to deescalate immediately

if that were to happen. So I haven't run any scenarios with

that specific threat in mind. But yeah,

it's always in the back of my mind that we'll get hacked in

some way that will either, you know, expose

some vulnerable information to the world or like

our customer information or, you know, in

this scenario, you know, affect our ability to dispatch or,

you know, send errant information out into the

universe to our customers. So it's always a threat, but

That's good. That's good. I've

got a couple more for you. And the first thing

is something that I've been thinking about, which

is, I wonder how

much you're thinking about or exploring ways that this

work expands beyond gun-specific detection.

You know, Zero Eyes doesn't have anything about guns

right in its name. And the idea of object detection

tied to, uh, warning systems with humans

in the loop connected to first responders seems to

me that it could scale to other situations. I

mean, like you, I even think about a fire, for

instance, in a, in a school building, the same way that, um,

your systems can tell first responders, which camera

flags, you know, an alert. it

would seem to be a very advanced fire alarm for you to be able to tell

someone there's a fire and this is exactly where it is. So

evacuate accordingly, right? Like that kind of thing. Other threats

beyond guns. I wonder if that's something that

From the very beginning, we've been exclusively focused on being the

best at one thing instead of mediocre at a bunch of different things. And

that's served us really well up until this point in the company. That

being said, our expertise in AI and

object detection lends itself really well to expanding into other

use cases like this. I

see it generally as we are a very high-value trigger for

an incident to start. And I would love to expand those

triggers to cover more incidents, things like perimeter

security and intrusion detection, health and safety, to

your point. There's

a lot of work being done on the retail side. with

loss prevention. So I see there's applications

across many verticals that touch the same customers that we do.

I first say that we're going to continue to be focused on guns, but start

to expand into some of these other areas that lend themselves

well. And ultimately, moving

into a future where vision transformers and

large language models are more easily accessible in real time, It

opens up a lot of possibilities. We could take this in a lot of different directions

and ultimately try to solve as many customer value problems as possible. But

yeah, very interested in intrusion detection, other types

of weapons like knives and just aggressive behavior in general,

things like that. And then ultimately we

want that trigger to initiate some sort of valuable response. Today,

that is us sending situational awareness to our customers that

they can respond, but we're still sending a person to deal with a

very dangerous situation. And so I would love to initiate

additional response like, you know,

locking down doors or sending a drone to verify the

incident and just have additional eyes on, I think is a

very likely future for us. If you can imagine a first

responder showing up to the scene after there's been a drone there for five minutes,

showing them what exactly is going on, they have the ultimate situational

awareness that they wouldn't have otherwise. And given

the dangerous nature of first responder jobs and security jobs,

I think it's completely natural that at some point in the

future, there will be some sort of response

Yeah, there's a lot of places you can take it, I guess, when you start thinking

about it like that. Sure. You know, drones, automated

security. I

guess you almost see the beginnings of like a Robocop type thing.

It's definitely hard to differentiate and diverge your thinking between

what is sci-fi and what is real life. And we get questions about

that all the time. Total Recall type analysis

where privacy is a thing of the past. It's

something we're sensitive to, and as much as possible, I want to

avoid privacy issues, because I care about that personally, myself.

But at the same time, there's so much possibility on the response

side to incorporate drones

and robots, autonomous response, into areas where otherwise

Right. I even think specifically for something like

the Hurt Locker, like if that could be done completely with robots, that

would be good. Setting people in the suits to diffuse

these things is crazy. But that also raises an

interesting point that I always think

is interesting because like we've mentioned

several times, you guys have been around for several years before

the kind of boom in AI that we're dealing with now.

And I wonder, I guess, how,

if it got easier to talk

to clients about what your offering is, when

AI became much more in

the public vernacular, And if in a weird way

that also made it a little bit more difficult, because a

lot of it is tied up, as you mentioned, right? With, with

people from outside the industry who are, have a hard time sifting through

fiction and, and, and the hype with what's actually happening.

Yeah. Our biggest challenge in the early days was convincing people that

it wasn't snake oil. And so, I mean,

we ran so many demos. At that point, just trying

to prove to people that we could actually detect guns. And then

they said, well, you know, you're showing us you being

detected with a gun. I want to hold the gun and be detected to

really prove it out. And so a lot of it has been proving out

that. AI security of

the past was maybe embellished

on the sales side to an extent that it caused customers to

think that AI wasn't capable

of performing in real time on security cameras. And we spent the

last seven, eight years trying to change that opinion a

little bit. And then going

forward with the

emergence of large language models and vision transformers

and some of the

common topics around them like copyright infringement, potential

lawsuits, these large language models using

public data in some way. It's

caused a lot of questions about how we collect our data, and that's been one

of our strong selling points to the customer is that All

of this data is organic to ZeroEyes. There's

no risk of us infringing on

any copyrights or using any public data. It's all stuff

that we've meticulously developed in-house

and have scrubbed to maintain the highest quality possible. So

I would say those are probably the two areas that we see overlaps with

Yeah. Yeah, you operate at

the intersection in an interesting way.

It's different from a lot of other AI companies that

I talk to. But my last

point, I guess, to leave off here, is,

you know, I wish you guys weren't needed. And it's

interesting, you start the company,

this team of veterans, in

the wake of a devastating mass shooting and,

you know, that's a problem and a crisis

that hasn't really abated too

much. And so here, you know,

where you sit with a technological solution, and

this kind of goal to, through

detection and better response times to mitigate violence

on the ground, whether that's someone has

an airsoft gun, or whether that's, you know, a mass shooting

might be about to happen, we have to get over there. You're

the idea of the kind of technological solution to a problem that bleeds

beyond technology. I wonder what you think about that and what

other levers should be considered or

if the reality of the world we live in is one where it's

like technology is

kind of the best thing we have left to

I also wish we weren't needed. I think about that almost

on a daily basis. As a company, we follow gun violence throughout

the country really closely. And so I

see on a daily basis the news reports of shootings and

gun violence that happen all over the country. And yeah,

when people come up to me and they say, how's business doing? I

say, it's good. Unfortunately, people keep

committing gun violence. And there's probably

deeper issues at play there. But

when I think about our position with the customer, we're

providing a layer of security. And so when we go

into a new customer, They trust us based

on our track record and expansion to basically

be an expert to them about how their security should look. And

so we're able to really provide this great feedback where, you know,

maybe a customer is struggling with their camera system and

it doesn't even make sense to buy ZeroEyes until their camera system

is upgraded. And so we're happy to make that recommendation to

them because At the end of the day, our performance won't be as

great on older cameras, lower resolution cameras. And

it matters to us that our customers have the highest security posture

that they possibly could. So we're in this really awesome

kind of expert position for our customers. And all

good security comes in layers. We're just one piece of

it. And we're trying to address that gap of being

that first early warning sign to communicate to

customers when there's a visible gun, when there's a weapon that's brandished on

their physical site. Yeah,

going forward into the future, I see a huge possibility for us to

be really ingrained with the customer to be that expert and to

provide that

knowledge that's needed to understand

how to respond to one of these incidents. Hopefully, the vast majority of our

customers will never experience one of these incidents, but with

the prevalence of gun violence and how it's expanding, that's

becoming more and more likely. And it's not as easy for customers to

just say, you know, that's not going to happen to me. They

have to be prepared in some way, and we're in this awesome position to help

them prepare. So I think we're just one

layer in their broader physical security, but we're a

critical layer at this point. I hope I answered your question. That

Thanks. Yeah. But from that perspective, the future

is exciting. But, you know, again, dark

Yeah, we as a company, we have to do well

in order to do good. That's one of our kind of principles. In

other words, we have to be profitable in order to expand to more cameras

so that we can cover enough cameras in order to fulfill our

mission, which is stopping gun violence. And

so we live and breathe it

at a daily basis here at Zero Eyes. And

when people come in to work every day, they're singularly focused

on that mission of ending gun violence. It's a beautiful place to work

because of that, but it does come with Um,

It's a heavy mission. Well, Tim, I appreciate you, uh,

letting me steal you away from it for, for a little bit, um,

and walking me through what you do. Uh, so thank you. Thank