Construction Vehicle Safety System

The rear view camera and surround view system just displays the camera stream on the screen, but this technology detects surrounding objects and alerts the driver of hazards.

This technology uses a deep learning based object detection model to identify objects. It is implemented as an embedded system with AI accelerators and machine learning frameworks.

Reason for start-up?

Our members have long been devoted to the research and development of High Performance Computing systems(High-performance Server & Storage), and the AI (Artificial Intelligence) era has come to be. So, we want to help the AI (artificial intelligence) market evolve more actively by providing cheap and high performance infrastructure through our technology and products. We also want to do something that will help the artificial intelligence market to prosper and benefit the society by providing an efficient infrastructure to harmonize the technologies and requirements of various fields.

HASPER Co., Ltd. is a place where such people can gather together to achieve harmony.

Meetaap Parallax


High Performance Computing Infrastructure Design Technology
– High-performance system design technology using Multi-GPU
– High-speed I / O interface design technology
– Edge Computing System design technology
– PCI-Express Switching design technology

Application technology optimization using FPGA
-Logic Architecture implementation technology using FPGA
-Co-processor implementation technology through various logic operations

Deep learning and machine learning based application solution development
– Smart Farm development
– Digital Signage development

Clustering-based high-performance computing deployment technology


Computing system comprising multi high speed processing devices and control module (2018. 11)

Meetaap Parallax

Problems & Solutions

  • In the meantime, in terms of research areas such as AI, Machin Learning, Deep Learning and service operations, we had to pay for expensive infrastructure costs for quick results analysis or we had to assume that the results of research, development and services could be late.
  • Reduce system operating costs by maximizing utilization of system space and lower cost of high-performance infrastructure


  • Delayed analysis time due to use of low-cost infrastructure
    >Opportunity cost loss
  • For fast analysis results, you have to pay for expensive infrastructure costs


  • High Performance & Low Price System
    >Reduced analysis time
    >System cost is lower than comparable performance systems
  • Scalability & Low Power consumption
    >Reduced system operation cost


  • Low Price system
  • Reduced system operation cost
  • More accurate is better
  • Faster is better
  • Computationally Intensive
  • Communication Intensive

Harsper Line up


HARSPER Product Features

High Performance Computing System

GPGPU based, High-speed operation system for AI, Machine Learning, Deep Learning, etc.


– 8 GPGPU modules can be installed in 1U System
– 16 GPGPU modules can be installed in 2U System

Low Power Consumption

System design using low-power modules reduces power consumption compared to comparable systems

High-speed multiprocessor and interface expansion technology

Patented technology that utilizes PCI-E expansion technology to effectively configure multiple GPGPU devices

Meetaap Parallax

Case by

Authoritatively streamline high standards in information before standards compliant benefits. Proactively e-enable standards compliant technology without value-added internal or organic sources.

Case by

Authoritatively streamline high standards in information before standards compliant benefits. Proactively e-enable standards compliant technology without value-added internal or organic sources.

Meetaap Parallax

Application of Machine Learning

Amazon relies on machine learning to improve customer experience in key business areas such as product recommendations, alternatives, fraud detection, metadata validation, and knowledge acquisition.

  • Demand Focecasting
  • Vendor Lead Time Prediction
  • Pricing
  • PakagingSubstitute Prediction
  • Product Recommendation
  • Product Search
  • Visual Search
  • Product Ads
  • Shopping Adice
  • Customer Problem Detection
  • Fraud Detection
  • Predictive Help
  • Seller Search & Crawling
  • Browse-Node Classfication
  • Meta-data vaildation
  • Review Analysis
  • Hazmat Prediction
  • Named-Entity Extraction
  • XRay
  • Plaglarism Detection
  • Echa Speech Recognition
  • Knowledge Acquision

Market Requirement

Real-time Data collection

Real-time data analysis

Heavy data handling

Server & storage investment based on scale-out approach

Cloud platform for big data analysis needs better H/W performance & consistency (Scale-up)

Large Data handling with performance

Consistent data handling capacity

Data security

Large volume of data gathering from manufacturing lines



Service Provider

Biological Sciences










Deep & Machine learning

Autonomous driving

Big data




Medical & Healthcare




Government Research institute

Education research institute

Genome analysis vendor

Data Center

Broadcast stations

Meteorological office




Telecommunication company

Results of Machine Learning

What did you actually achieve using Machine Learning?


ML adopters say the technology has led to more extensive data analysis and insights

More extensive data analysis :: more insights 45%
Faster data analysis : increased speed to insight 35%
Enhanced R&D capabiliities (next-gen products) 35%
Improved efficiency of intemal processes (operations) 30%
Better understanding of customer 27%
Competotove advantage 26%
Cost reduction 23%
Improved efficiency of external processes 19%

In a study by Google and MIT, we looked at the actual results for organizations that currently use machine learning, 45% said that machine learning led to extensive data analysis and insight, 35% answered that faster data analysis is possible.


Benchmarks with Tensorflow (Mempire™ MG140-G1)

This benchmark is a measure of the GPU’s computing power using TensorFlow, deep learning library provided by google.


Benchmarks with 3DMark (Mempire™ Edge-G1)

This benchmark is a measure of the GPU’s computing power using 3DMark.


Contact us


519-ho, 42, Changeop-ro, Sujeong-gu, Seongnam-si, Gyeonggi-do, Republic of Korea

Contact Numbers

TEL +82-31-778-7117