Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
The technology industry is in the limelight these days, but not for good reasons. It is the controversy surrounding US Immigration and Customs Enforcement and log into the Flock camera network monitoring people, and home camera ring drawing criticism by building a new system that would allow police to ask homeowners to describe their neighborhoods, there is currently a huge debate about security, privacy, and who can see.
But conflict doesn’t eliminate markets, and the continued advancement of visual language models has only added wind to the sails of companies that are building new ways to help companies manage what’s happening in their environments.
According to Matan Goldner, co-founder and CEO of the video management startup ContourThe movement around the topic is so important that he says his company is selective about the customers he can sell to. That may not make much sense as a two-year-old business, but Goldner says he can afford it because Connor already has several major government and public-listed clients, one of which is Singapore’s Central Narcotics Bureau.
“The fact that we have big customers allows us to choose them and be in control (…) We are in control of what they are using, how they can be used, and we can choose what we think is good and, of course, acceptable. We use all our judgment, and we make decisions based on the specific customers that we are good (to work with) because we know how they will use it,” Goldner told TechCrunch.
That pull has helped Conntour with more than just selection. Investors take note: The startup recently raised $7 million in seed funding from General Catalyst, Y Combinator, SV Angel, and Liquid 2 Ventures.
Goldner said the cycle was closed within 72 hours. “I think I scheduled about 90 meetings in eight days, and three days later – we started on Monday and by Wednesday afternoon, we were done,” he said.
Regardless, Conntour may be right on the mark, especially given how powerful AI tools in this area have become. The company’s videos use AI models to allow security guards to query the cameras using natural language to find any object, person, or event in the video, in real time — a Google-like search engine designed specifically for security video feeds. It can review and detect threats automatically based on pre-set rules, as well as ground-based information.
Unlike systems based on objects that rely on predefined definitions or parameters to identify specific objects, processes or behaviors, Conntour says its system uses natural and authentic languages, which makes it easy to adapt and use. A user can ask, “Find examples of a person wearing shoes passing a bag in a lobby,” and Conntour’s system will quickly search through all the recorded videos or video clips to bring up relevant results.
And because the platform burns in AI models, users can ask questions about the video and get answers in voice, followed by relevant video feeds, as well as generating reports about what happened.
The company’s selling point, however, is its problem. Goldner explained that the platform is different from other AI video search services because it is designed to run smoothly on systems with thousands of cameras. In fact, he said, Conntour’s system can manage up to 50 cameras from a single consumer GPU like Nvidia’s RTX 4090.
The company does this by using a number of models and logical methods, then determining the models and processes that the algorithm should use for each query in order to require the lowest possible computing power to give users the best results.
Contour says its system can be deployed entirely on-premises, in the cloud, or a combination of both. It can connect to many security systems already in use, or it can serve as a standalone monitoring platform.
But there’s been a long-standing problem with the video industry: Reviews are only as good as what’s filmed. It’s hard to tell much from a photo of a poorly lit parking lot taken with a low-end camera with a dirty lens, for example.
Goldner says Conntour hedges this inevitability by providing a reliable win with its search results. If the camera’s feed source is not good enough, the machine will return results with low confidence.
Going forward, Goldner says the biggest technology challenge it will solve is bringing the full LLM experience to its systems while maintaining efficiency.
“We have two things that we want to do at the same time, and they are contradictory. On the one hand, we want to provide the flexibility of natural language, the LLM style, to allow you to ask anything. And on the other hand there is power, so we want to use very few things, because again, preparing (thousands) of food is just madness. This contradiction, our barrier, the biggest barrier is what we work hard on, it is very difficult. to solve.”