Please help contribute to the Reddit categorization project here

    [–] In 1999, my mother took a $12/hour H-1B visa job so we could escape gun violence in South Africa. I'm now the CEO of a startup that uses AI to save lives during active shooter incidents. AMA! sonnytai 1 points ago in IAmA

    You start by networking to get meetings with the investors that you want to target. At the early stages, it will probably be angel investors. Find someone that knows them to give you an introduction. Then you should have your presentation ready to go, and be prepared to pitch them on it and answer any questions.

    Very few investors will invest on the first meeting. It often takes months, and you need to continue going back to them and informing them of your latest traction metrics (e.g. product progress, customer progress etc.)

    Happy to get on a call if you'd like to chat about it.

    [–] Axon Accelerate 2019 - Anyone Attending? sonnytai 4 points ago in ProtectAndServe

    Haha, if you come, I'll buy you beers!

    [–] In 1999, my mother took a $12/hour H-1B visa job so we could escape gun violence in South Africa. I'm now the CEO of a startup that uses AI to save lives during active shooter incidents. AMA! sonnytai 1 points ago in IAmA

    Thanks so much for asking. I'm often worried about her, but I fly back to Texas to visit her every month (she lives there now - better weather). I'm flying back at 6am, so I'm pulling an all-nighter tonight. No point sleeping by now!

    [–] In 1999, my mother took a $12/hour H-1B visa job so we could escape gun violence in South Africa. I'm now the CEO of a startup that uses AI to save lives during active shooter incidents. AMA! sonnytai 1 points ago in IAmA

    Athena is pretty good at PR, but they don't have a single employee with a strong back end engineering or data science background (check their LinkedIn page). I'd love to be proven wrong, but it is unlikely that their product works as advertised.

    Our core differentiator? We think that we are now the best in the world at using computer vision to identify gun threats - we got there by singularly focusing on this problem and painstakingly building a proprietary data set without trying to do everything else at once.

    [–] In 1999, my mother took a $12/hour H-1B visa job so we could escape gun violence in South Africa. I'm now the CEO of a startup that uses AI to save lives during active shooter incidents. AMA! sonnytai 1 points ago in IAmA

    Awesome question. We have heard talk of Avigilon building such a capability, but we have not seen it. Even if they have a gun-recognition system that works well, based on what we've heard, it appears that it would be a "smart camera" system for Avigilon customers - which is great, but what about the remaining 95%+ of customers that don't use the Avigilon VMS? Our solution is VMS agnostic.

    [–] In 1999, my mother took a $12/hour H-1B visa job so we could escape gun violence in South Africa. I'm now the CEO of a startup that uses AI to save lives during active shooter incidents. AMA! sonnytai 0 points ago in IAmA

    Thanks so much fr the encouragement - you hit the nail on the head. We are perfectly transparent that we are not a panacea. Addressing the gun violence epidemic will have to come from multiple angles - legislative, societal, and tech. We're just a team passionate about the issue and want to build something that can make an impact in the best way we know how.

    Most institutions have IP cameras - that is, networked cameras that stream to a video management system - basically an on-site windows box with enterprise software that manages all of your security camera feeds.

    We stream frames at 3 frames per second, and we compress frames to 608x608 before processing in the cloud, which means that it is not exceptionally onerous on a customer's network bandwidth.

    It is certainly possible though in large deployments that a customer would have to increase its bandwidth to handle the additional traffic. However, we are also building the capability to not process any frames unless a camera detects motion, which will dramatically reduce bandwidth and computation requirements.

    [–] In 1999, my mother took a $12/hour H-1B visa job so we could escape gun violence in South Africa. I'm now the CEO of a startup that uses AI to save lives during active shooter incidents. AMA! sonnytai 1 points ago in IAmA

    We can provide real-time visual intelligence that an acoustic-based system can't, and it's far cheaper because we integrate with existing security cameras - not significant hardware installation required.

    Obviously, one limitation would be that our system only works in areas covered by the line-of-sight of your existing cameras.

    [–] In 1999, my mother took a $12/hour H-1B visa job so we could escape gun violence in South Africa. I'm now the CEO of a startup that uses AI to save lives during active shooter incidents. AMA! sonnytai 2 points ago in IAmA

    My mother was 40 years old at the time, has a university degree, and is trilingual (Mandarin, English, Spanish). She was hired as a corporate procurement manager. At $18-2019 dollars-per-hour, you're right, we weren't starving, but things were not easy for a single mom raising two children.

    [–] In 1999, my mother took a $12/hour H-1B visa job so we could escape gun violence in South Africa. I'm now the CEO of a startup that uses AI to save lives during active shooter incidents. AMA! sonnytai 2 points ago in IAmA

    What is the use case of using computer vision to detect guns?

    We connect with existing security camera infrastructure, sample frames every second through the video management system, and run it through our computer vision model on AWS.

    Does your system notify security and the police?

    Currently internal security, but as we grow, we can build an operations center that can handle law enforcement alerts for the customer.

    Does it quantify the classifications and give first responders information such as the locations and dispositions of the shooter(s)?

    The cameras are all labeled, e.g. "north hallway camera, building 12, classroom 100", so the human-in-the-loop provides this information to police dispatchers, along with the observed threat (e.g. tall Asian man with gray t-shirt, blue jeans, wielding a rifle).

    Is there any plans to integrate it with security such that it could attempt to control doors to isolate shooters, perhaps also clear a path for potential victims to exit the building or unlock doors so they can get to a safe shelter in place area?

    This is something that would be engineered by a security integrator, but that's definitely possible.

    I imagine there are applications beyond this that you have thought of. I am only thinking of schools but of course this technology should be deployed to all public buildings and gun free zones. Is there any way to use this in non-gun-free zones?

    Will it couple the sight of a gun in a legal carry area perhaps with fleeing behavior?

    These two seem related - we only identify "gun in hand", which means that for legal carry areas, we're not looking for concealed or holstered weapons. If a firearm has been drawn, whether for self-defense or criminal reasons, it should be sufficient cause for an alert.