Who_we_are

What_we_do

Technology

Customers

Partners

 

 

MIND™ is a rugged (MIL-STD-810G) COTS device based on TPU (Tensor Processing Unit) and the RBF-RCE neuro-chip NeuroMem® by General-Vision.

MIND™ was designed to operate as an external coprocessor serving several on-board computers.

We design and build custom MIND™ prototypes for mission critical applications and harsh environments (MIL-STD-810G) using different hardware and software technologies.

We research new materials and construction technologies for specific uses in extreme environments. We create customized prototypes of accelerators for Machine Learning based on CPU, FPGA, TPU, GPU, Neuromem®.

RHPRD™  (Radiation Hardened Pattern Recognition Device ™) are based on three main critical factors:

1) Advanced proprietary space radiation shielding technology (NASAT™) in COTS solutions (TPU, Neuromem® VLSI).

2) Radiation Hardened components: algorithmic parallelism techniques (MYTHOS™) on Radiation Hardened CPUs like the BAE SYSTEMS RAD750® (Von Neumann Single Core) and Neuromem® IP (SIMD architecture) on Radiation Hardened FPGA.

3) NHTMR™ (Neuromorphic Hardware Triple Modular Redundancy™).

 

 

 

External coprocessor serving four on-board computers

 

 

 

 

  

 

4NME MODEL

 

·       2.68 TOPS

·       2.68 TOPS

·       Performs high-speed ML learning and inference

·       2.68 TOPS

·       RBF architecture with RCE learning algorithm.

·       Continuously learning classifier with L1 and L-SUP Norm.

·       KNN inference capability.

·       2000 neurons.

·       512K synapses.

·       Up to 128 different neural networks.

·       Explainable inference.

·       Ultra-low power (223 GOPS per watt).

·       Linux and Windows support.

4TPU MODEL

 

·       Performs high-speed ML inferencing.

·       The on-board Edge TPU coprocessor is capable of performing 4 trillion operations (tera-operations) per second (TOPS), using 0.5 watts for each TOPS (2 TOPS per watt). For example, it can execute state-of-the-art mobile vision models such as MobileNet v2 at almost 400 FPS, in a power efficient manner.

·       Supports all major platforms.

·       Debian Linux, macOS, or Windows 10.

·       Supports TensorFlow Lite.

·       No need to build models from the ground up.

ENVIRONMENTAL FEATURES (MIL-STD-810G)

 

·       Armoured enclosure

·       Sealed IP-68 (water / oil / dust)

·       G-Shock  proof

·       Vibration proof

·       Military temperature range

·       EMI-Shield

·       RH-PRD™ (Radiation Hardened)

·       Customizable MIL-DTL connectors

 

 

 ©2024_Luca_Marchese_All_Rights_Reserved

 

 

 

Aerospace_&_Defence_Machine_Learning_Company

VAT:_IT0267070992

NATO_CAGE_CODE:_AK845

Email:_luca.marchese@synaptics.org

Contacts_and_Social_Media