Reface and Prisma Co-Founders Collaborate to Advance On-Device AI Solutions
In the evolving landscape of artificial intelligence, the focus has shifted from cloud-based systems to enhancing on-device capabilities. A noteworthy player in this space is Mirai, a London-based startup founded by Dima Shvets and Alexey Moiseenkov, which is dedicated to optimizing AI model performance on smartphones and laptops. With a $10 million seed investment led by Uncork Capital, Mirai is on a mission to empower developers with efficient tools for AI integration.
Shvets, known for co-founding the popular face-swapping app Reface, and Moiseenkov, a key figure behind the viral AI filter app Prisma, recognized a strong demand among developers for improved cost efficiency and functionality in mobile AI applications. Their collaborative efforts aim to create a streamlined pipeline that facilitates complex tasks directly on devices.
Mirai is currently developing a powerful inference engine designed specifically for Apple Silicon, which promises to enhance on-device processing speeds significantly. The forthcoming software development kit (SDK) will allow developers to easily integrate this engine into their applications with minimal coding. Shvets articulated the vision, highlighting the intention to provide an experience akin to integrating a payment platform: “You basically go to our platform, integrate the key, and start working with summarization, classification, or whatever your use case is.”
Constructed using Rust, the engine can reportedly accelerate model generation speeds by up to 37% while maintaining output quality. Initially focused on improving text and voice processing, Mirai has plans to expand into visual applications. The team is currently collaborating with leading model providers to tailor their solutions for edge use and is engaging various chip manufacturers for future enhancements.
Additionally, Mirai aims to develop benchmarks for on-device performance, facilitating assessments for model creators. While Shvets acknowledges that some AI tasks will still require cloud processing, the startup is also working on an orchestration layer to manage requests that exceed on-device capabilities.
As the demand for efficient cloud inference grows, investors are recognizing the potential of startups like Mirai. Andy McLoughlin, managing partner at Uncork Capital, emphasizes that changes in the economics of AI workloads will drive a shift towards edge processing solutions, positioning Mirai advantageously in a competitive market.
Mirai’s seed funding round attracted various notable investors, enhancing its potential for growth and innovation in the on-device AI space.
