CN

Company Profile

InfiScale: Empowering AI Chip  Ecosystem Development

Shanghai InfiScale Electronic Technology Co., Ltd,   established in December 2024, is dedicated to fostering an software ecosystem for Artificial Intelligence (AI) systems and to offering low-cost, high-performance, and scalable computing resources.  Built around its foundational software toolchain, the company expedites the market entry of edge chips and facilitates the low-cost deployment of AI applications. 

ESTABLISH

Edge Servers

Industrial visual inspection (PCB defect detection), Retail analytics (people counting, product recognition)

MISSION

Smart Home

Localized face/voice recognition (person detection), Environmental perception (abnormal sound detection: glass breaking, smoke alarm)

POSITION

Medical Devices

Wearable Health Monitoring (Arrhythmia detection) Portable diagnostic devices (supporting real-time assisted diagnosis for ultrasound, X-ray images)

Product Overview

Edge Inference Chip

Technological Advantages

Vertical integration between algorithms, systems, and chips with compiler optimizations

E2E adaptation to large model characteristics with higher throughput and low power consumption.

Continuous performance improvement over model inference.

Application Scenarios

Large model inference on edge personal computers and mobile devices

Reduced model computation costs based on quantization and sparsity techniques

Cost-effective computing power requirements

Cloud/Edge Inference Optimization Tool

Technological Advantages

Holistic optimization between AI models, systems, and hardware

Standardized software foundation across various chip brands and models.

Seamless improvement of model inference performance.

Application Scenarios:

High-performance inference demands by reusing code between data centers and edge devices

Compatible inference-side deployment for models trained on Nvidia and other vendors

Efficient model deployment with optimized computing power reduction

Auto-Generation and Translation Tools

Technological Advantages:

Automated Generation

Performance Guarantee

Cost Advantage

Application Scenarios:

Provides efficient CUDA-compatible operator translation, Auto-generates operators
with support domestic chip support, Emphases inference development efficiency

Ensures high-performance operator translation for different chip architectures,
Guarantees the high standards and operational efficiency of AI applications

Reduces development cycles and costs Accelerates domestic chip time-to-market

Edge Inference Chip

Cloud/Edge Inference Optimization Tool

Auto-Generation and Translation Tools

Contact Us

Address

Building Q, Pudong Software Park, No. 66 Boxia Road, Pudong New Area, Shanghai

Mail-box

marketing@infiscale-tech.com

Tel

021-68650181