Mirai Secures $10M to Build On-Device AI Layer
Mirai, a startup focused on edge AI, has raised $10 million in seed funding to build an on-device AI capability layer. The company aims to make local model inference more efficient and accessible for applications where privacy, low latency, and reduced cost are primary concerns.
- The founding team's background at Reface and Prisma provides them with extensive experience in shipping consumer applications with on-device AI, having navigated the complexities of performance optimization for millions of users. - Mirai's proprietary, Rust-based inference engine is initially optimized for Apple Silicon, demonstrating up to 37% faster generation speeds compared to Apple's own MLX framework, a key performance indicator for developers building responsive applications. - The startup's strategy includes a hybrid approach, offering an orchestration layer that allows developers to route AI workloads between the device and the cloud. This enables enterprises to manage costs and performance by processing sensitive or low-latency tasks locally while leveraging cloud resources for more intensive computations. - For agentic AI, on-device processing provides two critical advantages: ultra-low latency for real-time decision-making and direct access to local data and context on the device, enabling more personalized and autonomous workflows without compromising privacy. - The investment was led by Uncork Capital, whose partner Andy McLoughlin has a history of investing in developer tools and B2B software. The round also includes notable angel investors such as the co-founder of Snowflake and the former CTO of Stripe, signaling strong confidence in Mirai's potential to provide essential infrastructure for the next wave of AI applications. - Mirai is addressing a significant economic driver for enterprises by aiming to reduce the high costs associated with cloud-based AI inference. As companies scale their AI features, the "pay-per-request" model of cloud services becomes a major expense, making on-device processing a more cost-effective solution. - The company plans to release on-device performance benchmarks, which will allow enterprise and independent developers to evaluate how various models perform locally. This move towards transparency will help shape API and model selection strategies for companies building on-device AI capabilities. - Governance of on-device AI introduces new challenges for enterprises, such as managing model versioning across a fleet of devices and ensuring compliance in decentralized environments. Mirai's SDK-based approach will be a critical point of implementation for such governance frameworks.
