- About Us
We didn't pivot to Edge AI.
We grew into it.
From embedded systems to Edge AI
— a natural progression
Founded in 2004, embedUR systems built embedded software for leading technology companies across telecom, networking, consumer electronics, and industrial systems. That deep expertise in resource-constrained hardware evolved into Edge AI as inference moved from the cloud to the device for better privacy, lower latency, and greater efficiency.
ModelNova was created to turn more than 22 years of embedded engineering into practical Edge AI. Through the Model Zoo, Fusion Studio, licensable production-ready models, and custom model creation, we help product teams build and deploy Edge AI without starting from scratch. ModelNova is an independent Edge AI company born out of embedUR Systems, with its own roadmap, brand, and commercial identity.
embedUR systems founded in Fremont, California. Focused on embedded software for enterprise networking, telecom, and connected devices.
Edge AI initiative formulated within embedUR systems. Engineering teams begin porting and optimizing neural networks for MCUs.
embedUR ships its first production AI pipeline: network monitoring running on tens of millions of devices worldwide. Still in production today, powering deployments at carriers around the world.
embedUR systems joins the TinyML Foundation (now EDGE AI FOUNDATION) as a Platinum Strategic Partner. [reference]
embedUR partners with Synaptics at Embedded World in Nuremberg, demonstrating real-time object recognition (YOLO) running on a battery-powered Synaptics Astra device — breakthrough AI performance on a low-power edge processor. [reference]
ModelNova Zoo launches at the TinyML Summit in Milan: a model zoo of pre-trained, edge-optimized AI models for vision, speech, and sound built to take developers from idea to proof-of-concept in days, not months. [Reference]
ModelNova Zoo powers EDGE AI LABS – enabling academia, hobbyists, and industry leaders to collaborate and curate a forward-thinking Model Zoo for Edge AI. [reference]
ModelNova runs the first build of what would become NovaEyeD, its production face recognition pipeline, on STMicroelectronics' STM32N6. [reference]
embedUR brings Edge AI to Embedded World: a booth at the industry's largest embedded show, with 7+ live Edge AI demos running on Infineon, NXP, STMicroelectronics, and Synaptics silicon. [Reference 1 & Reference 2]
learn more
Fusion Studio enters beta with initial support for Raspberry Pi: a desktop development environment that takes models from proof-of-concept to production retraining on your own data, optimizing for your target, then benchmarking, validating, and deploying in a single tool. [Reference]
ModelNova brings Fusion Studio into the Arm Ethos-U NPU ecosystem, starting with Alif Semiconductor's Ensemble MCUs (Arm Ethos-U55 and Ethos-U85) — its first step beyond Raspberry Pi into production Edge AI silicon. [Reference]
ModelNova showcases NovaEyeD at Embedded World, running production-grade face recognition across Infineon, STMicroelectronics, and Alif Semiconductor silicon — one model, proven on multiple platforms. [Reference]
embedUR’s ModelNova and STMicroelectronics host a joint webinar on NovaEyeD, demonstrating production-grade face recognition running on the STM32N6. [Reference]
ModelNova launches as a company in its own right - an independent Edge AI company born out of embedUR Systems, with its own brand, roadmap, and commercial identity.
Mission
To simplify and accelerate Edge AI development – giving developers the models, tools, and IP to go from concept to deployed device in days, not months.
- Why ModelNova
Key Differentiators
We build for real hardware
512 KB of RAM, 0.1 W of power, GOPS-class NPUs — these are the constraints we design within every day. We don’t adapt cloud AI frameworks down to the edge. We built for the edge from the start.
Production, not research
Our models go into commercial products, not academic papers. NovaEyeD is already shipping in real devices, validated on production silicon. When we say production-ready, we mean it has already shipped.
The full Edge AI journey in one place
Most Edge AI tooling solves a single slice of the problem. ModelNova covers the whole path: a Model Zoo to start from, Fusion Studio to retrain, optimize, benchmark and deploy, and licensable production-ready models to ship – all from one company, under one roof.
Hardware-agnostic by design
We don’t favor any single silicon vendor. Our models and tools are validated across Alif Semiconductor, Infineon, NXP, Renesas, Silicon Labs, and STMicroelectronics. You bring the silicon – we optimize for it.
Two decades of embedded, now on AI
ModelNova was born out of embedUR systems, bringing more than 22 years of shipping software on constrained hardware. We’re a new company, but not a first-timer – we’ve been solving the hard part of edge AI since before it had the name.
Company Facts
Headquarters
San Ramon, California, USA
Chennai, India
Engineering Centre
India — multiple locations within the embedUR organization
Models deployed
NovaEyeD in commercial production in home appliances and sensing applications
Silicon partners
8 active co-development partnerships: Arm, Alif, Atmosic, CEVA, Infineon, Nordic, NXP, STMicro
Our Locations
- Headquarters
California, USA
- Engineering Centre