Cloud Native Computing Foundation Adds 46 New Members

Accelerating Hyperparameter Tuning with Container-Level GPU Virtualization

Top 15 Korean AI Startups to watch for – Best of 2019

South Korea looks to own startups for growth, here are the top 5 in 2019

Backend.AI in the News

 ![](https://cdn.lablup.com/Lablup_GTC_24_a1eb755914.png) 

[Mar 4, 2024] Lablup is set to participate as a silver sponsor at NVIDIA's premier AI conference, GTC 2024, taking place from March 18th to 21st at the San Jose Convention Center, USA. This in-person event, the first of its kind in five years, is anticipated to be the largest tech gathering in the GPU sector, with over 300,000 participants joining both online and offline.

At the conference, Lablup will present on two cutting-edge topics within AI technology. The first presentation, titled "From Idea To Crowd: Manipulating Local LLMs At Scale," will be conducted by Lablup's CEO, Jeongkyu Shin, and CTO, Joongi Kim. This session will introduce how Backend.AI leverages resource-saving and model compression technologies such as quantization and QLoRA, along with the integration of vLLM and TensorRT-LLM, to efficiently manipulate large-scale local LLMs. This approach aims to optimize AI model performance while minimizing resource usage.

The second presentation, "Personalized Generative AI: Operation and Fine-Tuning in Household Form Factors," scheduled for March 20th, will be led by CTO Kim Joongi. It will demonstrate how to develop and service customized generative AI using cloud and PC GPUs, presenting new possibilities for integrating AI technology more closely into everyday life through personalized AI.

Furthermore, Lablup will showcase Backend.AI Enterprise at booth #1233, the only NVIDIA DGX-Ready software in the APAC region. Backend.AI is an AI infrastructure operation platform that maximizes the performance of NVIDIA DGX systems and other GPU-accelerated computing resources, enhancing GPU usability without altering the user environment. At the booth, Lablup will showcase FastTrack, an MLOps platform that streamlines and automates AI development, customizing foundation models for various industries and enabling real-time fine-tuning for personalized chatbots.

Following MWC, Lablup aims to strengthen cooperation with global partners such as NVIDIA, Dell, AMAX, and Weka at GTC, paving the way for a significant global expansion.

Lablup to Showcase the Future of AI Innovation at NVIDIA GTC 2024

 ![](https://cdn.lablup.com/MWC_Amax_744f5049a8.jpeg) 

[Barcelona, Feb. 26, 2024] Lablup announced on the 26th that it has held a ceremony to sign an agreement with global HPC solution company 'AMAX' at 'MWC 2024', the world's largest mobile exhibition held in Barcelona, Spain. The agreement aims to strengthen business development cooperation and expand business opportunities between the two companies.

The ceremony was attended by Lablup CEO Shin Jeongkyu and AMAX Senior Vice President Bohr Tsao. Lablup, a company specializing in AI research, development, and service platforms, and AMAX, an HPC solution company, plan to accelerate the construction of AI supercomputing infrastructure and provide GPU integrated solutions to research institutes and data centers worldwide.

The two companies have been discussing collaboration since co-hosting the "Optimising AI Conference" in London last November. At this event, technical experts and business leaders in the AI field, along with Lablup, presented cases across the AI industry, including telecommunications, healthcare, and finance, based on Lablup's UK market entry case, and had networking time.

Bohr Tsao, Senior Vice President of AMAX, said, "The current problem with IDC is not only power consumption but also heat control. We have introduced a configuration that installs a separate liquid cooling rack along with a calculation rack, which allows us to build the optimal heat control infrastructure even in the current IDC environment." He added, "We look forward to supplying a global AI infrastructure platform by combining AMAX's infrastructure design and operation experience with Lablup's Backend.AI."

Shin Jeongkyu, CEO of Lablup, said, "Lablup, which provides the world's best platform for increasingly complex AI infrastructure operation and optimization, and AMAX, which provides high-performance computing hardware platforms of various scales, are expected to create synergy." He expressed his ambition, saying, "Through this agreement with AMAX, we will provide an AI total platform targeting the high-performance computing demand in Europe and North America, allowing more people around the world to enjoy AI innovation through Backend.AI more closely."

Lablup Signs Global Agreement with AMAX for AI Infrastructure Platform Supply

![](https://cdn.lablup.com/Gen_AI_Dasktop_01_02538c826c.png)

Lablup announced its participation in the world's largest mobile exhibition, 'MWC 2024', held in Barcelona, Spain, from February 26th to 29th.

The major topic of this year's MWC 2024 is AI. Telecommunication companies are eagerly announcing their investments in the latest AI-related technologies with the aim of joining the global trend of the edge cloud market. Lablup will introduce how telecommunication companies can easily, quickly, and cost-effectively research, develop, and serve AI on the edge clouds, which are small computational resources set up at base stations. Furthermore, it plans to enter the global edge cloud market through partnerships with global enterprises.

At this exhibition, Lablup is demonstrating "GenAI Desktop," an easy-to-use AI application integration platform for the general public. This platform is a tool that helps users easily understand and utilize AI. In addition, through Backend.AI FastTrack, attendees can directly demo the chatbot solution "Talkativot", developed by Lablup through fine-tuning large language models to fit specific needs. FastTrack is a B2B solution aimed at corporations and research institutions, simplifying and automating AI development through an MLOps platform. This platform manages all processes from data analysis, sorting, and processing to AI training, evaluation, and service deployment rapidly and effectively.

Lablup's CEO, Jeongkyu Shin, stated, "Lablup aims to 'Make AI Scalable' this year and will leap to become a platform company leading the global AI market through edge cloud-based AI solutions and generative AI application platforms." He added, "Lablup, participating as a silver sponsor at the NVIDIA GTC 2024 to be held in March, will lead the latest AI trends not only in the domestic market but also in the global AI

Lablup to Showcase AI Edge Cloud Solutions and Generative AI Application Platform at MWC 2024

 ![](https://cdn.lablup.com/Gen_AI_Desktop_bf679fb88c.png) 

AI research and development platform company Lablup announced that will participate in the world's largest consumer electronics and IT trade show, CES 2024, in Las Vegas, USA. on January 9.

At the exhibition, Lablup will unveil "GenAI Desktop," a generative AI application integration platform for general users, and will introduce a demonstration of using generative AI applications directly from anywhere. "GenAI Desktop" is a platform that allows applications that utilize AI models to run like desktop apps anywhere, and is characterized by the fact that users can select and run the services they need like an app, without having to have detailed knowledge of AI operations. One of GenAI Desktop's native apps, "Talkativot," which allows users to fine-tune large language models that are the hottest theme around the world and use them as chatbots. It will be demonstrated at Lablup's booth (#63301-51).

Lablup will also demonstrate Backend.AI FastTrack, an MLOps platform that simplifies and automates generative AI development as a B2B solution for enterprises and research labs. FastTrack quickly and effectively manages every step of the pipeline, from analyzing, classifying, and processing the data that makes up an AI project to training, evaluating, and deploying services. It makes all AI tasks replicable and sharable, so repeated processes can be easily reused or shared. Backend.AI FastTrack, evolved for the era of generative AI, is also available at the booth for running AI services at scale on GPU hyperscaler clouds.

Lablup, which is participating in CES for the second consecutive year, has been invited to participate in the K-Tech Forum@CES 2024 Korea pavilion, organized by KOTRA, as a startup panelist in the field of AI, and will share its experience in collaborating with global companies such as NVIDIA, PureStorage, NetApp, and Dell as technology partners and exporting to the UK and Brazil.

"Lablup has been constantly thinking and innovating on different ways to make AI easy, safe, and convenient for the world to use," said Jeongkyu Shin, CEO of Lablup. "This CES will be the beginning of introducing our solution to the world wide public on how to use AI in a friendly way." He added.

Lablup Unveils GenAI Application Integration Platform "GenAI Desktop" at CES

![](https://cdn.lablup.com/ai_relib_award_9d1147b73b.JPG)

The 'AI Reliability and Quality Award' is the first award system in Korea that focuses on AI reliability and quality, and aims to spread awareness of AI reliability and support the promotion of excellent domestic AI products and services by selecting and awarding AI products and services (including data) that have secured reliability.

Backend.AI is an all-in-one platform that integrates the entire process of researching and developing AI and creating services, and is an enterprise solution that specializes in the fields of super large-scale AI, high-performance computing, and supercomputing. It is installed on on-premises servers and cloud environments, and is responsible for smooth distributed execution and management of deep learning model design and training processes. Based on Linux container and VM technology, it securely distributes and executes AI and high-performance computing code on various platforms, and automates complex AI development processes that mobilize thousands of computers with GUI and MLOps functions without complex settings.

Backend.AI integrates the latest storage and AI acceleration hardware technologies on top of its proprietary orchestrator and scheduler, along with its own patented GPU partitioned virtualization technology that significantly reduces the cost of GPU usage. This enables the company to achieve industry-leading performance, efficiently reducing the cost of massive and costly AI training and the rapidly growing number of AI services.

Backend.AI is open-sourced so that anyone can develop AI without barriers, and Backend.AI Enterprise, an enterprise version with many features for enterprises and institutions, is recognized as an AI platform that integrates high performance and convenience, and is used by more than 70 enterprises and institutions, including Korean conglomerates such as Samsung Electronics, KT, and LG Electronics, as well as the Bank of Korea, the Health Insurance Review and Assessment Service, and GIST. It was also selected as the only DGX-Ready software in the Asia-Pacific region that is optimized to operate NVIDIA's AI supercomputer, DGX, and is currently exporting Backend.AI Enterprise to Latin America and Europe.

Founded in 2015, Lablup secured a Series A investment of 10.5 billion won earlier this year. "Backend.AI technically solves various reliability problems that occur in the process of AI development and operation.

"Backend.AI is helping many companies to create highly reliable and high-quality AI. We are pleased that Backend.AI itself has been recognized for achieving the reliability and quality requirements of various AI developments." "This award will be a great opportunity for Backend.AI to secure customers and users, and will help us to be more active in the global AI market," said Shin.

Backend.AI Receives AI Reliability and Quality Validation to Demonstrate Excellence

![Backend.AI package.jpg](https://cdn.lablup.com/Backend_AI_package_1276cda2a8.jpg)

  Lablup announced today that will participate in "SC23 (November 12-17)" at the Colorado Convention Center (Denver) in the United States. The conference is an event for high-performance computing, networking, and storage analytics, and attracts top domestic and international companies such as AMD, ARM, AWS, IBM, Intel, NVIDIA, and SK hynix.

  Lablup will introduce its platform, Backend.AI, for constructing a super-giant AI ecosystem at this year's SC23. Not only does Backend.AI support a variety of GPUs, including NVIDIA CUDA and AMD ROCm, it also offers optimized features for various NPUs and AI semiconductors. It supports Google TPU, Graphcore, Rebellion Atom, and Furiosa Warboy, among others, along with optimization features. Utilizing a GPU-NPU integrated pipeline to train AI models on GPUs and serve them on AI semiconductors, it achieves both the highest performance and the lowest cost.

  Furthermore, Backend.AI is the first globally to implement GPU Direct Storage technology in a container cluster, providing tremendous data I/O speeds by directly connecting the GPU and network storage. This has led to partnerships with the top three in the high-speed storage market: PureStorage, NetApp, and Dell EMC. In addition, Backend.AI was the first and only one selected for the NVIDIA DGX-Ready program in the Asia-Pacific region and was chosen as a partner of the NVIDIA AI Accelerated Program along with VMWare, RedHat, and Canonical.

  At SC23, Backend.AI will be demonstrating LLM services that can be directly developed and operated using this technology. It plans to announce that it can develop and serve LLM services in an on-premises environment through Backend.AI's automation system, FastTrack, by fine-tuning a variety of large public language models, including the Llama2 model disclosed by Meta, for various application areas. They will also showcase VisuTale, a multimodal demo that creates stories and generates images based on user-provided photos, subsequently creating a picture book.

 "We are excited to introduce Backend.AI, which is optimized for high-performance computing, along with various supercomputer solutions, to supercomputing researchers around the world. We will continue to strive to achieve performance and cost innovations and overwhelming convenience through Backend.AI in technology-leading fields that require huge amounts of computational resources," said Jungkyu Shin, CEO of Rabbleup.

Lablup introduces the hyperscaled AI service built from Backend.AI at SC23

![2023-08-llm-chatbot-screenshot.png](https://cdn.lablup.com/2023_08_llm_chatbot_screenshot_e548d4462f.png)

Lablup wowed developers at PyCon Korea 2023, held August 11-13, with features and demos that automate the process of creating, fine-tuning, and servicing large language models with Backend.AI, its AI R&D and services platform software.

In the demo, various open-source large language models, including the recently released Llama2 model from Meta, were fine-tuned for Korean using Backend.AI's automation system, FastTrack. The chatbot, which was created on the fly with the Llama2 model fine-tuned in 12 hours, was serviced during the event via Lablup's NVIDIA H100 cloud farm with a dedicated UI, attracting the attention of many participants.

Backend.AI provides leading features in various AI development and service fields based on ▲ maximizing GPU cluster utilization, ▲ intuitive user experience and management, ▲ low-cost/low-latency AI model and service technology, and ▲ support for various AI semiconductors. With the unveiling of its large-scale language model development, fine-tuning, and service automation capabilities, Lablup has demonstrated that Backend.AI can be utilized for full-scale generative AI applications such as task automation, research, and programming code development.

"We have dramatically simplified the process of developing and integrating various services based on large language models into enterprises, institutions, research institutes, and organizations that require security. We are confident that any organization can use Backend.AI to easily fine-tune and operate their own unique large language model to meet their needs."

Currently in beta testing with select Backend.AI customers, the big language model integration will be available to on-premises customers as part of the general availability of Backend.AI in September. The service will also be on display at KSC2023 (August 21-23, The K Hotel Seoul), Korea's largest supercomputing conference and exhibition.

> Automatically translated from Korean Article

Lablup to Debut Massive Language Model Integration Development and Services at PyCon Korea 2023

Lablup Inc. (CEO Shin Jeong-kyu), an AI research, development, and service platform company, announced that it has finalized a **Series A investment** of **KRW 10.5 billion**. LB Investment, IBK, K2 Investment Partners, and Daesung Startup Investment participated in the round.

With the recent ChatGPT craze increasing interest in generative AI services, investors recognized the value of Lablup, which develops and sells Backend.AI, a platform that enables companies to efficiently operate GPUs, AI semiconductors, and data required for AI training and services.

Currently, Backend.AI is used in a variety of AI development and service fields, including the development of large language models and generative AI for pictures, text, music, etc. and the use of AI education platforms, based on maximizing GPU cluster utilization, intuitive user experience and management, low-cost/low-latency AI model and service technology, and support for various AI semiconductors.

Lablup's Backend.AI is used by large Korean companies such as Samsung Electronics, KT, LG Electronics, CJ, and more than 70 other companies and organizations, including the Bank of Korea, the Korea Health Insurance Agency, and GIST, and has been selected as NVIDIA's DGX-Ready software, the only total AIOps platform solution for AI development and services in the Asia Pacific region.

"As interest in AI R&D and services continues to grow globally, so does the demand for solutions that can provide much-needed resources for GPU utilization and cost savings," said Jungkyu Shin, CEO of Lablup. "This investment will enable us to play a leading role in the AI platform solution market, both domestically and globally."

> Automatically translated from Korean Article

Lablup, an AI research, development, and services platform, raises 10.5B KRW (Approx. 80K USD) in Series A funding

Lablup Inc. (CEO Shin Jeong-kyu) today announced that it will introduce Backend.AI R3 at the Mobile World Congress (MWC) 2023 in Barcelona, Spain.

![2023_02_mwc_1.jpg](https://cdn.lablup.com/2023_02_mwc_1_2380bd8e59.jpg)

The company will introduce Backend.AI, a platform that enables companies to efficiently operate GPUs, AI semiconductors, and data for AI training and services, to the European and global telecommunications industry and mobile-related companies at the show.

Backend.AI is a total AIOps platform solution for artificial intelligence (AI) development and services that is the only DGX-Ready software from NVIDIA in the Asia-Pacific region. It enables efficient use of the compute and power required to train very large AI models. The Backend.AI platform automates the AI research and service development process and provides the necessary resources within an organization to enable anyone to create and service AI without time and cost constraints. Recently, Backend.AI's MLOps platform, FastTrack, and Forklift services were integrated into an enterprise solution. This enables faster AI service development cycles, reducing the cost and time of AI development for various organizations.

"MWC 2023 is a place to see the reality of digital transformation that is revolutionizing existing industries," said Shin Jeong-kyu, CEO of Lablup. "New technologies and business models are emerging through the convergence of innovation networks and AI." "Lablup is preparing AI model services based on Backend.AI that respond to various needs, from domestic telecommunications companies and giant language models to generation models, and MWC 2023 will be the place to introduce how global companies that want to adopt AI can achieve convenience and cost flexibility together."

> Automatically translated from Korean Article

Lablup to Participate in MWC 2023, Introduce AI R&D Platform to Increase Cost Competitiveness of Ultra-Large AI Model Services Like ChatGPT

*   Providing the best technology for container-based AI/ML model development and operational efficiency
*   Expected synergies between the two companies...from implementing GPU environment for AI/HPC to continuous operation management

Hyosung Information Systems (CEO Yang Jeong-kyu), a data infrastructure and solutions company, and Lablup Inc. (CEO Shin Jeong-kyu), a company specializing in hyper-scale AI development platforms, announced on the XXth that they have signed a memorandum of understanding (MOU) to expand the domestic AI (artificial intelligence), ML (machine learning), and HPC (high-performance computing) markets and strengthen collaboration, including joint sales and marketing.

Successful AI projects require advanced strategies for efficient utilization of GPU computing resources, effective storage and performance maintenance of growing data, easy development of AI models, and continuous operation management.

Hyosung Information Systems is strengthening its pursuit of the GPU-based AI platform business with Lablup, which is recognized globally for its AI development platform and GPU partitioned virtualization technology. Based on proven products and various project know-how, the two companies will provide the best infrastructure environment for customers' AI projects, from GPU environment implementation to operation management.

Hyosung Information Systems simplifies complex AI tasks for companies through its 'AI Integrated Platform' that combines performance-optimized GPU server infrastructure and AI model operation management system. It provides integrated proposals from GPU-based data computation to large data storage and automated management and operation using storage optimized for high-performance computing, actively supporting customers' data-based digital transformation.

Lablup provides Backend.AI, an all-in-one enterprise cluster platform for AI development and services. Backend.AI provides all the functions required in the era of AI development and service, including self-developed container-based GPU partitioned virtualization technology, deep learning model distributed training and ultra-large model development platform technology, and GPU-based ultra-low cost/ultra-low latency AI model service technology. It is currently being used by leading domestic and international companies and organizations, including Samsung Electronics, KT, LG CNS, SK, Lotte Communications, CJ, and others.

"Based on our technology partnerships such as NVIDIA's AI Accelerated Program and DGX-Ready Software Program, as well as our proven technology through large-scale enterprise and institutional customers at home and abroad, we will enhance our synergies for our customers' successful AI projects." "From implementing a GPU infrastructure environment to developing effective AI models based on containers and optimizing service management, our customers will be able to accelerate the introduction of AI services and create business results based on AIOps," said Jungkyu Shin, CEO of Lablup.

"Based on the data and AI-related technologies that both companies have accumulated in Korea, we will provide the best infrastructure environment for customers who are experiencing difficulties with GPU virtualization and resource management," said Yang Jeong-kyu, CEO of Hyosung Information Systems.

> About Lablup Inc. ([https://www.lablup.com](https://www.lablup.com))
> 
> Founded in 2015,  a company specializing in hyper-scale AI development platforms, has been recognized globally for its AI development and service platform Backend.AI, which is the first in the Asia-Pacific region to be certified as NVIDIA DGX-Ready Software.
> 
> Backend.AI ([www.backend.ai](https://www.backend.ai)) enables effective management of AI/ML/HPC from R&D to business services to AI service inference and delivery through one consistent platform and is utilized by various customers such as Samsung Electronics, KT, LG CNS, SK, Lotte Communications, SK, Bank of Korea, and ETRI.
> 
> > Automatically translated from Korean

Hyosung Information Systems and Lablup Strengthen Market Penetration of GPU-based (Hyper-scale) AI Platforms

During the 2022 GTC keynote, NVIDIA introduced the AI Accelerated program partners. Among the more than 100 companies was Lablup, a Korean company that developed Backend.AI, a hyperscale AI R&D and service platform.

NVIDIA's AI Accelerated program is for AI-enabled application developers and solution partners who leverage the NVIDIA AI platform to deliver faster, more efficient, and better results for their customers. The program includes some of the world's best-known companies, including Adobe, VMware, and Canonical.

Having already been certified as the first and only DGX-ready software from NVIDIA in the Asia-Pacific region last summer, Backend.ai's inclusion in the AI Accelerated program is a recognition of its global competitiveness and excellence in the MLOps and AI services market.

Backend.ai is a total platform solution for cloud and on-premises that builds an AIOps environment for AI development and service, automating the entire process from AI R&D to service through a standardized platform and providing the necessary resources through an already owned AI cluster or cloud, enabling anyone to create and service AI without time and cost constraints.

Users can conveniently develop AI models on laptops, editors, and professional terminals, and developers can integrate Backend.ai into their own software and platforms based on a powerful SDK. In addition, operators can intuitively perform individual, group, and domain-specific management through the control panel and view various real-time statistics through the dashboard.

"Backend.ai's selection as a partner in NVIDIA's AI Accelerated program, a leader in AI technology, will enable the MLOps ecosystem to further accelerate and lead the AI development process for all users at home and abroad through more advanced hyperscale AIOps," said Jungkyu Shin, CEO of Rableup.

> Automatically translated from Korean Article

Lablup Announces Backend.AI as a Partner in NVIDIA's AI Accelerator Program

today announced that Lablup Inc. has been selected as a global technology partner in the Technolpgy Alliance Partners (TAP) program with Pure Storage, a leading provider of all-flash storage platforms.

The TAP program recognizes Technolpgy Alliance Partners who collaborate with Pure Storage to build and deploy pre-tested joint solutions, including AI, analytics, cloud, and modern data protection, to deliver and build better businesses.

Rableup's Backend.AI is a solution that efficiently leverages hardware, software, and multi-cloud-sourced cloud and data center compute resources to run and manage large-scale computational workloads, including large-scale AI model development and high-performance computing, by optimizing the performance of accelerated computing, including GPUs on Pure Storage FlashBlade, to efficiently manage compute resources to maximize GPU usage without changing the user experience.
(Reference: https://www.purestorage.com/partners/technology-alliance-partners/lablup.html)

"The development of AI research is no longer just for research, but for real-world applications, and the infrastructure for such development is becoming large-scale and sophisticated," said Jungmook Kim, Chief Operating Officer of Rableup. "By becoming a TAP with Pure Storage, we are leading the global market in providing highly usable hardware to build such infrastructure and support for optimal solutions that maximize that hardware to further accelerate the AI development process and optimize everything from deployment, usability, and utilization of AI infrastructure."

> Automatically translated from Korean

Lablup Partners with Pure Storage to Expand Global Footprint

Today Lablup Inc. announced that Backend.AI, a solution that efficiently leverages cloud and data center compute resources to run and manage large-scale computational workloads, including large-scale artificial intelligence (AI) model development and high-performance computing, is the first in Asia Pacific to be certified as NVIDIA DGX-Ready Software.

NVIDIA DGX-Ready Software is a validated, enterprise-grade software solution that is fully tested and certified for use on clusters of NVIDIA DGX systems, and Backend.AI is the first company in Asia Pacific to be certified.

Backend.AI optimizes the performance of NVIDIA DGX systems and accelerated computing, including NVIDIA GPUs, to efficiently manage compute resources to maximize GPU utilization without changing the user experience. In addition, Backend.AI has its own patented container-based GPU partitioned virtualization technology, which allows users to partition expensive GPUs into multiple GPUs as needed to maximize GPU usage.   
(Reference: [https://www.backend.ai/partners/nvidia)](https://www.backend.ai/partners/nvidia))

"Backend.AI has been successfully supporting all types of AI platform customers, from educational institutions to Hyperscale AI Infra," said Jungkyu Shin, CEO of Lablup Inc. We are delighted to be the first in Asia-Pacific to be certified as NVIDIA DGX-ready software, and we look forward to helping our users, both in Korea and abroad, not only accelerate their AI development process, but also optimize their deployment, usability, and utilization of their AI infrastructure."

NVIDIA DGX systems accelerate large-scale AI initiatives to help organizations solve key challenges in healthcare, manufacturing, commerce, science, and more. The NVIDIA DGX-ready software program helps customers succeed by validating enterprise software solutions like Rableup's Backend.AI that streamline workflows and optimize utilization across AI clusters.

> Automatically translated from Korean

lablup.com

lablup.com

News and Press

Backend.AI in the News

Press Releases