ThinCI’s unique approach towards deep learning is based on harnessing the graph technology, a unique framework to store and process graphs.
TECHNOLOGY
HIGHLY PARALLEL, GRAPH STREAMING PROCESSOR
Our systems use graph structures for semantic queries with nodes, edges and properties to represent and store data. The GSP is suited for on-chip task graph execution. At the heart of the GSP architecture lie multiple levels of parallelism - task, thread, data and instruction level. Integrated inside the GSP’s processor fabric is “fine-grained thread scheduling” which is aware of data dependencies, and it dispatches threads when resources are available and dependencies are verified. The massively parallel architecture is designed to process data-intensive workloads across deep learning, AI and computer vision.
FULLY PROGRAMMABLE GRAPH COMPUTING COMPILER
We make application development easy by offering industry-standard APIs and extended built-in libraries for vision processing and machine learning. The layered Graph Computing Compiler (GCC) captures the “intent” of the graph problems that need to be solved from complex and existing code and processes it in parallel — in a streaming manner. The comprehensive SDK supports machine-learning framework such as TensorFlow, Caffe, Torch and also OpenVX + Open CL, C/C++ language kernels.
INTEGRATED SOLUTION
Supports all types of neural networks commonly used
INTELLIGENT BEHAVIOUR
The GSP understands complex data dependencies and flow and executes in sync
BETTER EFFICIENCY
Better system performance at a lower system power consumption.
EASE OF PROGRAMMING
Supports popular frameworks to enable software teams ship their code faster.
We are hiring exceptional professionals to join our technology and business teams across locations.