Mirai Raises $10M for On-Device AI Layer
Bay Area startup Mirai has raised $10 million in seed funding to build an on-device AI "capability layer." The company's goal is to make local AI inference more efficient and accessible for developers. The focus on edge AI could influence the development of privacy-preserving on-chain analytics and trading agents.
- The $10 million seed round was led by Uncork Capital, with participation from notable angel investors including the CEO of Dreamer David Singleton, a partner at Y Combinator Francois Chaubard, and a co-founder of Snowflake Marcin Żukowski. - Mirai was co-founded by Dima Shvets and Alexey Moiseenkov, the entrepreneurs behind the popular AI-powered consumer apps Reface and Prisma. Their experience with these viral applications, which together amassed over 200 million downloads, highlighted the need for more efficient on-device AI processing. - The company's core technology is a proprietary inference engine built with the Rust programming language and initially optimized for Apple Silicon hardware. Mirai claims this engine can increase the speed of AI model generation by up to 37% without compromising the quality of the output. - Beyond performance gains, Mirai's "capability layer" is designed to simplify the process for developers to integrate on-device AI, offering a developer-friendly SDK that can be implemented with just a few lines of code. - The company is developing an orchestration layer that will manage workloads between the device and the cloud, allowing for a hybrid approach where less demanding tasks are handled locally for privacy and speed, while more complex requests are sent to the cloud. - Future plans include expanding support beyond Apple Silicon to other hardware platforms and moving from text and voice to include vision and multimodal AI models.
