Inspector Cloud, an automated auditing startup powered by some of Skoltech’s great computer vision minds, has won a coveted spot in a New York accelerator – an honor that comes accompanied by a grant of upwards of $130,000.

This victory didn’t come easy; Inspector Cloud was up against 600 competitors from 30 different countries, vying for 13 spots in this year’s Starta Accelorator program, which aims to build bridges for Russian and Eastern European companies wishing to launch their businesses stateside.

Inspector Cloud was developed to relieve many of the labor and resource burdens traditionally associated with monitoring a store’s inventory.

“We came up with the idea about a year ago that retail needs a solution to monitoring the presence of goods on shelves,” said Alexander Berenov, Inspector Cloud’s Chief Executive Officer (CEO).

“We had many talks with prospective partners and clients and determined that actually they needed this technology and were searching for a Russian vendor. We then decided to focus intensely on developing a solution,” he said.

Inventory monitoring is necessary for reasons ranging from annual accounting requirements to meeting customer demand and responding to a lack of customer interest in certain products.

Traditionally, the process has required employees to manually count items on a store’s shelves and in its storage facilities, and subsequently to compare these figures with those in the company’s accounting records.

Many businesses require an ongoing general monitoring of supplies, as well as a comprehensive inventory counting process at least once a year. As these comprehensive counts typically have to take place after hours, the result is long hours for employees and costly overtime pay for management.

Inspector Cloud decimates the time and energy involved in the traditional process by creating an automated alternative to the manual process. Specifically, using the startup’s software, shop employees can photograph goods on a store’s shelf, and then enter these photos into a system that relies on neural networks to categorize and count the goods.

According to the company’s website, its algorithm is able to properly identify 96% of products based on photos; the 4% that can’t initially be identified are then used to train the neural network in order to prevent subsequent mistakes.

Based on these results, Inspector Cloud is able to provide retailers with real time audits, and to notify store management within three minutes of detecting abnormalities.

In addition to revolutionizing the auditing process, the company claims that its services can boost sales by giving retailers keen insights into which products tend to fly off the shelves, which ones tend to gather dust and what correlating factors may have a bearing on either scenario.

Inspector Cloud also works with fast-moving consumer goods (FMCG) companies like beauty empire L’Oréal and chewing gum giant Wrigley, which have an interest in monitoring the share of shelf space their products occupy in stores as compared with their competitors, as well as the movement of their goods in certain stores.

“We offer retailers a free license to use our services, and in exchange we take the data we retrieve from the stores’ usage of our service and then sell it to the FMCG brands. This is a novel business model in this market, and it’s one of our main distinguishing points,” Berenov said.

Skoltech’s Pavel Boyko, a senior research scientist with the Center for Computational and Data-Intensive Science and Engineering (CDISE), serves as Chief Technology Officer (CTO), while Professor Viktor Lempitsky, who leads Skoltech’s Computer Vision Group, serves as a scientific advisor.

Boyko explained that his role encompasses a vast range of responsibilities: “As CTO, I am responsible for product and technology development. As always with startups, this covers a broad area, ranging from listening to clients’ needs to implementing advanced deep learning algorithms to troubleshooting service continuity on weekends, just to name a few of my duties. I also manage a small but very talented R&D team. With time, my responsibilities have drifted from inventing technology demonstrators to satisfying clients on a daily basis with a bulletproof product.”

Boyko explained that his initial motivation to get involved with Inspector Cloud arose from his excitement over the potential convergence of offline retail with the latest breakthroughs in computer vision.

His early enthusiasm has borne fruit. “One year later, I still believe Inspector Cloud has offered the best product-market fit of any project I’ve taken on so far,” he said.

Lempitsky was drawn to the project based on its drive and promise: “I was attracted by the motivated and talented team, which I believed had a very good chance for success.”

He added that the team’s close affiliation with CDISE made the collaboration feel all the more natural.

Lempitsky’s goal in working on the project is to bolster the everyday use of computer vision technologies: “Automating retail inventory is a challenging use case for computer vision and some of the technology I have been working on in recent years with my students and collaborators. This technology includes recognition systems based on deep neural networks and large-scale visual search. Helping to implement such technology for a real-life task is obviously exciting.”

Now that Investor Cloud has secured a spot with the Starta Accelorator, Berenov is off to New York to participate in a 3.5-month intensive program aimed at arming his team with the strategies and funds they’ll need to penetrate the US market.