It’s kind of a metaverse camera: it captures the real in the virtual, so gamers can exchange clips of the best moments they’ve experienced during a video game session. For this, the Parisian start-up Powder has developed an artificial intelligence (AI) for recognizing images and emotions. The solution is concrete, based on the machine learning and on the deep learninganalyzes in particular the sound dimension. “When players start to react, when something loud happens, we detect patterns in the sound that we quantify. We then get an amplitude of emotions from which we can determine a strong emotional moment.”, explains Stanislas Coppin, co-founder and CEO of Powder. This recognition is now combined with algorithms “which also make it possible to recognize what is being said during a game”says the director. All to analyze, for example, the laughter or other emotions generated by the exploits of the players.
A growing international audience
To innovate in this sector, the AI department of the startup is working on the one hand to find new ways to read video and audio elements to extract information from them and on the other hand to create game analysis models. More than half of Powder’s 45 employees are developers, engineers or PhD students specializing in artificial intelligence.“We have 18 nationalities within our company”, says Stanislas Coppin. The challenge is to have a diverse team capable of creating an international product that meets the demands of different markets. In short, talents with diverse profiles (engineers, developers, designers, salespeople) bring more ideas together forever.
Especially now that the start-up continues to grow. Since April 2020, when the application was released on the App Store, there have been about three million downloads worldwide. Its clients are mainly located in Asia and also in the United States – countries where this Parisian start-up has been located since its inception. His ambition? “Our goal is to develop our international audience and then start monetizing, first across the Atlantic and in Asia, while continuing to develop our European market”, says Stanislas Coppin. For this, the company has already raised 20 million euros and is planning another round of financing in the coming months to consolidate its mission.
The AWS Cloud, a companion to machine learning
To realize its ambitions, Powder has relied on the cloud since its inception. Two main reasons for this. First, “simplicity of installation and management”analyzes Christian Navelot, co-founder and technical director of Powder. “To manage video, we need very large storage volumes and fast access times”he develops. “Without the cloud, we would have to manage our machines ourselves, which requires a lot of resources, dedicated network lines, a cooled space and considerable costs. This is all the more complicated in times of strong growth.”he believes.
But beyond storage, “Even in the cloud it is important to know how to use the right tools”, continues Christian Navelot. This is exactly the element that motivated the choice for Amazon Web Services (AWS) that supports the nugget in its development. On the supplier side, a account manager and a solution architect are working with the start-up on better use of tools in the cloud. “Cost is a real issue for a growing startup like ours. To optimize it, AWS supports us in the regular assessment of our needs »summarizes the technical director of Powder.
“In particular, we use Amazon Sage Maker, the cloud-based machine learning platform that provides many AI and Machine Learning services. It allows us to use certain tools, for example for the training and industrialization of machine learning models. So we bring some models to life in parallel to be able to refine them, especially in the case of the release of new seasons of games”provides information about this manager.
Finally, Powder was able to unlock $100,000 in credits per year through the Activate program. The key for the startup is the ability to run multiple trials on the AWS platform. Another way to speed up his climb.