Dima Shevts, co-founder of global hit Reface, the most downloaded face-swapping app in the world, joins entrepreneur Alexey Moiseenkov. Collectively they are behind the helm of a groundbreaking new organization called Mirai. London-based Mirai would like to change how artificial intelligence (AI) models are developed and operated. Their main priority is improving performance on personal devices such as smartphones and laptops. The company prides itself on having a specialized technical staff of 14 technicians. Their goal is to create a robust framework that improves the performance of models locally on devices, reducing the necessity for costly cloud inference.
Since then, Reface has raised a $5 million seed round led by the perennial goldenboys of venture capital firm Andreessen Horowitz (a16z). With this cutting-edge app, users can easily change faces in videos and photos—with shocking realism. It quickly became a viral sensation. In much the same way, Shevts is the co-founder of Prisma, an app that exploded in popularity for its AI-based photo filters throughout much of the last decade. Through his experiences with these successful applications, Shevts has continuously explored the potential of AI and machine learning on personal devices.
While in London, Shevts and Moiseenkov discussed their vision for AI technologies that could perform complex tasks without needing constant internet connectivity. Their discussions were the impetus for the development of Mirai. This cutting-edge platform advances developers’ ability to perform sophisticated processes on mobile devices via simplified coding applications.
We launched the business with a specific purpose.
Onboarding developers
We wanted to make integrating with developers as painless and simple like Stripe as possible, something that could be handled in eight lines of code. Shevts elaborated. “You basically go to our platform, integrate the key, and start working with summarization, classification, or whatever your use case is.”
Mirai is already working to make its on-device AI even better. This push comes amid rising scrutiny over the cost of cloud inference. Active supporting venture capitalist, Andy McLoughlin, stressed how the current funding climate is biased towards companies that have made aggressive investments in cloud solutions. That’s an encouraging development, but he cautions that this model won’t be tenable in perpetuity.
“Considering how expensive the current cloud-based inference is, something has got to give,” explained McLoughlin. So far, VCs have been willing to bankroll the rocketship companies, pouring inordinate sums onto cloud inference. But it’s a reality that won’t endure indefinitely. Sooner or later, the public will begin to look at the economic fundamentals of these enterprises and realize that a reset must occur.
Mirai’s vision is part of a larger movement toward decentralizing AI processing. Increasingly, industries are seeing the promise of edge computing and on-device processing. Given these challenges, the kind of solutions that a company like Mirai can provide will only be in more demand. McLoughlin remarked on this shift: “It feels like every model maker will want to run part of their inference workloads at the edge, and Mirai feels very well positioned to capture this demand.”
Even still, shevts understands that not every AI workload is able to process natively on devices. The key thing he wants to emphasize about Mirai is that it was largely designed to solve complicated problems well, even in dynamic, mobile settings. Mirai is about capability but about practicality. This holistic dual approach makes it a leading contender in the rapidly evolving on-device AI solution landscape.