AuroraML raises $75M to launch real-time multimodal inference edge SDK

AI · 4 min read

AuroraML announced a $75M Series B today and the simultaneous launch of Aurora Edge, its developer SDK designed to deliver multimodal inference on edge platforms. The company says Aurora Edge can run a combined vision-and-audio model with tokenization and lightweight context memory within 50ms on modern NPU-equipped IoT processors, a claim that targets robotics, AR glasses, and smart cameras.

Investors in the round include Frontier Peak, StudioVentures, and an unnamed semiconductor partner that will bundle Aurora's runtime on upcoming modules. CEO Laila Moreno told SatisfiedUser the funds will accelerate compiler optimization, developer tooling, and support for additional NPUs and microcontrollers.

The SDK ships with sample models for keyword spotting plus gesture detection, an explainability dashboard for debugging multimodal decisions, and a privacy-first pipeline that keeps all raw sensor data on-device. AuroraML plans to open-source core runtime components while commercializing optimized model bundles and a licensing program for consumer device OEMs.