The minds behind viral photo apps Reface and Prisma are taking aim at one of AI's biggest infrastructure challenges. Mirai, a new startup focused on optimizing how AI models run directly on smartphones and laptops, just closed a $10 million seed round. The bet: as AI moves from the cloud to your pocket, the companies that solve the performance puzzle will power the next wave of intelligent devices.
The on-device AI revolution has a performance problem, and Mirai thinks it has the answer. The startup, founded by the entrepreneurial duo behind consumer hits Reface and Prisma, just secured $10 million in seed funding to build infrastructure that makes AI models run faster and more efficiently on smartphones, laptops, and other edge devices.
It's a pivot that makes sense when you look at where the AI market is headed. Apple is pushing Apple Intelligence across its entire product line. Google is embedding Gemini Nano directly into Pixel phones. Microsoft is racing to put AI into every Windows laptop with its Copilot+ PC initiative. But there's a catch - running sophisticated AI models on consumer hardware with limited power and processing capability is brutally difficult.
That's where Mirai comes in. The company is building tools that optimize model inference, the process of actually running AI models to generate predictions or outputs. According to the TechCrunch report, the team is leveraging their experience building consumer apps that processed millions of AI-powered photo and video transformations to tackle the infrastructure layer.
The founding team's track record is noteworthy. Reface, their face-swapping app, exploded to over 200 million downloads and became one of the most viral AI applications before the ChatGPT era. Prisma, their earlier venture, pioneered neural style transfer for mobile devices, turning photos into artworks using AI filters. Both apps ran into the same fundamental challenge - how do you deliver complex AI experiences without draining batteries or requiring cloud roundtrips that kill the user experience?
Now they're selling the solution to that problem. The $10 million seed round suggests investors see real demand for on-device AI optimization tools. As more companies try to ship AI features that work offline, run privately on-device, and don't hammer battery life, infrastructure tools like Mirai's become critical.
The timing aligns with broader industry shifts. Qualcomm is shipping chips specifically designed for on-device AI workloads. Nvidia is pushing its Jetson platform for edge AI deployment. And a growing ecosystem of model compression techniques, quantization methods, and specialized hardware is making it feasible to run increasingly sophisticated models locally.
But performance remains a massive bottleneck. Large language models that run smoothly in the cloud often crawl on consumer devices. Computer vision models that power augmented reality experiences can overheat phones. Voice assistants that process requests locally struggle with accuracy-speed tradeoffs. Mirai is betting that better inference optimization can unlock a new generation of AI-powered apps that don't compromise on experience.
The startup faces competition from established players like Google's MediaPipe and open-source frameworks like ONNX Runtime. But the founding team's consumer app DNA could give them an edge in understanding real-world performance constraints. Building for 200 million Reface users taught them exactly where models break down in production.
What makes this round particularly interesting is the signal it sends about infrastructure investing in the AI space. While much of the funding frenzy has focused on foundation model companies and AI application layers, the picks-and-shovels businesses solving deployment challenges are starting to attract serious capital. On-device inference is one of those unglamorous but essential problems that every AI product eventually confronts.
The $10 million will likely fund team expansion and partnerships with device manufacturers and app developers. For Mirai to succeed, they'll need to prove their optimization tools can deliver measurable improvements - faster inference times, lower power consumption, smaller model footprints - across diverse hardware and use cases.
Mirai's $10 million seed round marks a bet that the next frontier in AI isn't just smarter models - it's making those models actually work on the devices in our pockets. With major tech platforms racing to ship on-device intelligence and a founding team that's already delivered AI experiences to hundreds of millions of users, the startup is positioned at the intersection of a real infrastructure need and growing market momentum. Whether they can translate consumer app success into developer tools adoption will determine if this becomes the infrastructure play that powers the edge AI era, or just another well-funded attempt to solve one of computing's hardest optimization problems.
