1702ai Co-Founders Joachim Levy and Keren Levy-Bildi: Weapon Detection with AI
The team and company behind Samson, an AI gun detector video platform powered by deep learning algorithms.
Founded at the beginning of 2015, the company only raised a small round of
USD 100K in 2017 from a Geneva-based micro-vc. Led by a strong Swiss business culture, 1702ai focuses on generating revenues first rather than raising financial rounds. With an HQ in
Zürich, Switzerland and its own data center in
Abu Dhabi, the company developed SAMSON, a novel software that detects small weapons leveraging existing IT infrastructures. Aimed at IoT manufacturers and the public safety sector, the product is currently being deployed in three European smart cities.
How was the idea of 1702ai born? How would you describe its ultimate mission?
The idea was born after the Charlie Hebdo (Paris) and the Raanana Mall (Israel) attacks. 1702ai’s ultimate mission is to save lives by providing real-time situation awareness to first responders.
Joachim, Prior to founding 1702ai, you led your team to create the first VR (virtual reality) military exercises for the Swiss Military. How did that experience help you in founding and leading 1702ai?
I trust it was the first time the Swiss Government was using such an innovative tool as VR (Virtual Reality) to train their civil protection troops in 2017. This exercise taught us that when working with governments - patience, a well-thought R&D plan, and a fail free system are a must as there is no room for error.
Keren, being a female co-founder of a highly technical startup is still sadly a rarity. What advice would you give to female entrepreneurs, girls, and women, who are thinking about starting their own tech venture?
Ladies, if you are to start your own tech venture for power and fame, I would suggest to go a different route. However, if you are fine to be the "Neck" of the company rather than the "Head", then find a partner that completes you and is willing to be the face of the startup. Unfortunately, the world is not yet 100% ready for women leading tech ventures but we are getting there. Patience is key.
Because our first dataset was composed of 1702 images of real knife attacks. Yet our data set was entirely built in house, we do not use any open source data as much as you’d not use any free medicine you’d find on the streets.
For training we are huge fans of Amazon Sage Maker and, of course, we embrace Python as it allows quick in and quick out using great libraries.
To stop criminal activity, it is essential that your technology works in real-time. How has 1702ai achieved the real-time function?
Our technology is real-time and entirely runs on-premises using the latest Intel micro processing technologies widely used in the intelligent video analytics sector. We achieved real-time functions by having an excellent team that is fuelled by passion. Also, entirely removing the cloud from the equation allows for much faster processing. Additionally, it makes us much less prone to cyber attacks.
What role does AI/ML play in your technology?
It’s a big nut, as important as support, cyber and GDPR. Yet AI is by nature infused in our company. Everybody has to put their hands in AI so we can debate, dream, argue and breathe it. It also has a magical productivity multiplier when everyone does AI.
AI is by nature infused in our company. Everybody has to put their hands in AI so we can debate, dream, argue and breathe it.
My guess is that to err on the positive side, you have more false positives than false negatives. How do you address the false positives problem and strike the balance between detecting the weapons on all occasions and not bothering people with harmless behavior?
We have a human in the loop approach, we are an extra pair of eyes that never goes to sleep. That is when there is an alert we make sure humans are part of the decision process. Now imagine an AI that detects cancer cells. Well, it still needs a doctor before sending a patient to chemo. In some way, we are the same. Our AI would never alert the S.W.A.T. directly. That is why we strive to make our AI better day by day. Our AI is like cyber - it never sleeps and will never stop evolving.
Our AI is like cyber - it never sleeps and will never stop evolving.
For developing this threat detection system, 1702ai uses existing IT infrastructures and video cameras. How do you manage to integrate your technology so seamlessly given the endless number of CCTV providers?
As an image processing expert that was taught how to think by Apple and with an Oscar Nomination under my belt, I see problems from various angles. Additionally, I am winged by our brilliant COO and analyst Keren and our CTO Moshe that led IBM’s trends and technology for IBM’s world famous
R&D Zürich lab. By working together, we have a unique mindset that allows us to think entirely different. I used to handle studio productions in L.A. and work behind a million dollar camera, today I am still behind the camera, but a different one which is much smaller and cheaper, so it is actually a breeze.
I used to handle studio productions in L.A. and work behind a million dollar camera, today I am still behind the camera, but a different one which is much smaller and cheaper, so it is actually a breeze.
What was the most significant tech challenge that your engineers had to overcome while building SAMSON?
Weapon detection is so challenging that it was extremely hard to decide on what to focus first.
Was there a particular client story that you can share where 1702ai technology made a huge impact?
Our Jr. AI engineer was in a terrible knife attack that he witnessed, his motivation is off the chart, and by having him in our team we actually feel safer. On another note, as we are working with governments it is hard for us to share client stories due to confidentiality.
How did COVID-19 impact your business?
We became camels and learned how to save our water, therefore we only spend on necessities.
What do you think is next for AI + IoT in Security?
Bandwidth is a major factor that will have to be dealt with within the next years. Therefore the challenge will include solving great problems such as video compression.