The startup develops chip technology that integrates big data analytics and heterogeneous computing to enhance the functionality of the Internet of Things. This technology enables industries, such as automotive and healthcare, to incorporate artificial intelligence into their products and services, improving operational efficiency and decision-making capabilities.
Funding
$15.2M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Product
Problem
The increasing complexity and computational demands of artificial intelligence applications, particularly in edge computing and embedded systems, require specialized hardware solutions. Traditional general-purpose processors often lack the efficiency and performance needed to run AI inference models effectively in these environments. This results in higher power consumption, increased latency, and limited scalability for AI-enabled devices.
Solution
NovuMind offers the NovuTensor IP, a domain-specific architecture (DSA) designed from the ground up for AI inference acceleration. The patented architecture leverages tensors as the primitive data type, enabling significantly higher hardware efficiency and software simplicity compared to conventional approaches. This allows for power-efficient, cost-effective, and high-performance AI inference across a range of applications, from battery-powered embedded devices to high-throughput edge servers. The NovuTensor IP is highly scalable and customizable, allowing customers to optimize performance and power consumption for their specific AI workloads.
Target Audience
The primary target audience includes system-on-chip (SoC) designers and manufacturers in industries such as automotive, IoT, and consumer electronics who require high-performance, low-power AI inference capabilities for their products.
Features
- Domain-Specific Architecture (DSA) optimized for AI inference
- Patented tensor-based architecture for high compute efficiency
- Heterogeneous computing support for software simplicity
- Scalable IP configurations ranging from a few TOPS to PetaOPS
- Support for various foundry process nodes and libraries
- Demonstrated performance in object detection (YOLOv3 at full HD resolution at 30 FPS with two cores)
- Demonstrated performance in super-resolution (8K at 60 FPS with an 8-core configuration)