In the current landscape of artificial intelligence, discussions often revolve around expanding cloud infrastructure and massive data centers to support complex models. However, a dedicated team of 14 from London-based Mirai is shifting focus towards enhancing how AI models operate directly on smartphones and laptops.
Founded last year by Dima Shvets and Alexey Moiseenkov, Mirai has secured $10 million in seed funding led by Uncork Capital. Both founders bring valuable experience from their previous ventures, with Shvets co-founding the popular face-swapping application Reface, and Moiseenkov being the CEO and co-founder of the renowned AI filters app Prisma.
Prior to the rise of generative AI, both Shvets and Moiseenkov had been contemplating the potential of AI and machine learning on personal devices. During a meeting in London, they discussed the gap in the market for on-device AI solutions, as the focus remained predominantly on cloud-based applications. They recognized a significant opportunity to develop a pipeline that empowers users to execute complex tasks directly on their devices, leading to the inception of Mirai.
Mirai is currently creating a framework that enhances the performance of AI models on devices. The company has developed an inference engine specifically optimized for Apple Silicon, which aims to improve on-device processing capabilities. Their upcoming SDK will allow developers to integrate this technology into their applications with minimal coding, providing a seamless experience.
Shvets emphasized their vision for the company, stating, "We wanted to offer developers a straightforward integration experience, similar to what Stripe provides. A few lines of code can enable functionalities like summarization and classification, tailored to various use cases."
Built using Rust, Mirai's engine can reportedly enhance model generation speeds by up to 37% without compromising output quality. Currently, the focus is on optimizing text and voice functionalities, with plans to incorporate visual capabilities in the future. The team is also collaborating with leading model providers to adapt their technologies for edge deployment and is exploring partnerships with various chip manufacturers.
Moreover, Mirai intends to introduce on-device benchmarks to facilitate performance testing for model developers. While acknowledging that not all AI tasks can be executed on-device, the team is developing an orchestration layer to manage requests that require cloud processing.
Although the startup is not yet directly involved with application development, its engine has the potential to enhance on-device assistants, transcription services, translation tools, and chat applications.
Andy McLoughlin from Uncork Capital highlighted the changing landscape of AI, noting that as cloud inference costs rise, there will be a shift towards on-device solutions. He believes that Mirai is well-positioned to meet this emerging demand.
Mirai's seed funding round also attracted investments from notable individuals, including influential figures from various tech sectors.
