femtoAI Opens Developer Community to Build New AI Powered Products
FemtoAI, a Silicon Valley-based startup focused on AI silicon and compression, has opened its developer community, offering access to its full-stack platform. The company states its technology delivers a 10X lower power draw and 10X smaller memory footprint for AI-powered products without sacrificing accuracy.
The core of femtoAI's offering is a full-stack AI compression platform that leverages patented neural-network sparsity and advanced model compression. This approach is intended to enable significant reductions in power consumption and memory footprint, with customer examples like Legato and Marshall reportedly using the technology in smart glasses and audio devices.
The developer community provides access to a Sparse Processing Unit (SPU) chip evaluation kit, a suite of development tools, and options for model flexibility, including prebuilt models or the ability to integrate custom ones. The Sparsity Studio™ tool is offered to allow developers to measure the power and memory savings achieved through femtoAI's sparsity techniques.
Beyond hardware, femtoAI also applies advanced model compression to software. The company claims to have delivered open-source models, such as OpenAI Whisper, with a 10X reduction in memory, and developed custom solutions like local voice commands. This integrated hardware and software approach aims to expand the feasibility of deploying AI in devices with strict resource limitations.
Share this article
Related reading
6 stories
India backs Project VeerAI with Rs 130 crore debt to build home-grown AI vision chips
India is also backing home-grown AI vision chips, showing a similar national focus on AI hardware development.
BDO Unibank Powers Smarter, Next-Generation Payments With Finastra
China’s ‘city of the future’, Xiongan, hunting high-value tech firms: sources
The Anthropic researcher-turned-whistleblower says he was 'working to automate' himself
Can Microsoft Help Customers Cut Back on Claude?
