The startup develops ultra-energy-efficient neural network models that utilize nonlinear dynamics and space-time sparsity, inspired by biological neural networks, to enable large-scale processing at the edge. These models provide ultra-low latency, low power consumption, and enhanced user privacy for businesses requiring efficient user interfaces in complex processing tasks.
Funding
$16.3M 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
AI model deployment on edge devices is often constrained by the limited compute resources and power budgets of these devices. Large, complex models demand significant memory and processing power, hindering their use in applications like wearables, IoT devices, and other embedded systems.
Solution
Femtosense provides an embedded AI platform that enables efficient execution of large AI models on resource-constrained edge devices. The platform leverages a sparse processing unit (SPU) architecture, inspired by neuromorphic computing, to minimize power consumption and memory footprint. By exploiting sparsity in both AI models and hardware acceleration, Femtosense achieves significant performance gains compared to traditional processing approaches. This allows developers to deploy sophisticated AI capabilities, such as noise reduction, voice interface, and sound event detection, on devices with limited resources.
Target Audience
The primary target audience includes product developers and manufacturers of consumer electronics, IoT devices, and embedded systems who seek to integrate advanced AI capabilities while minimizing power consumption and cost.
Features
- Sparse Processing Unit (SPU) AI accelerator optimized for sparse neural networks
- Dual sparsity support: sparse weights and sparse activations for multiplicative efficiency gains
- Near-memory compute architecture to reduce data movement and improve throughput
- Scalable core design for flexible deployment across various applications and form factors
- Software Development Kit (SDK) with tools for sparse optimization, model performance simulation, and deployment
- Support for popular AI frameworks like PyTorch, TensorFlow, and JAX
- Ready-to-deploy AI models for applications such as AI noise reduction, voice interface, and sound event detection
- Low-power consumption, enabling always-on AI functionality on battery-powered devices