2025數位應用週 AI臺灣 創新無限

AFS (AI Foundry Service) POD Service offers an all-in-one, optimized integration of server systems, data center infrastructure, and AI software development capabilities—designed to meet the full spectrum of AI data center demands from governments, enterprises, and the global GPU market. Whether scaling AI resources or launching generative AI projects, our solutions accelerate deployment, optimize costs, and enhance operational efficiency.

TASC provides a comprehensive on-premises AI infrastructure solution, integrating ASUS servers, racks, power and cooling systems, data center-grade networking and storage, and a customized AI software platform. This platform supports compute resource management, application workflows, and billing system integration. Leveraging high-performance computing (HPC) technologies, we unlock the full potential of GPU-based processing, accelerating complex computations and data workloads to deliver unprecedented performance breakthroughs. A large-scale AI data center can be up and running in as fast as 8 weeks, with full enterprise-grade Service Level Agreement (SLA) support.

With over 7 years of experience and technical expertise in AI infrastructure deployment, TASC has helped build Taiwan’s largest AI supercomputers for both government and enterprise clients. Our systems have ranked among the global TOP500 high-performance computing systems, empowering national AI sovereignty, advancing smart city development, and accelerating industrial AI innovation and implementation.

The Formosa Foundation Model (FFM), independently developed by TASC, is the world’s first enterprise-grade large language model (LLM) optimized for Traditional Chinese. It is designed to advance sovereign AI development, strengthen data security, and drive industry transformation.

FFM is trained based on globally recognized open-source models and, within just two years, has iteratively launched multiple versions that retain native model capabilities while deeply understanding Traditional Chinese semantics and domain-specific knowledge. This includes models like Llama3.2-FFM, a vision-enabled multimodal model that significantly enhances the accuracy of Traditional Chinese image recognition, understanding, and Q&A.

FFM has been certified by Taiwan’s Artificial Intelligence Evaluation Center (AIEC) under the Ministry of Digital Affairs, ensuring secure and localized deployment. It supports both cloud and on-premises deployment and offers flexible commercial licensing. Through integration with the AI Foundry Service (AFS)—a one-stop platform for model optimization and inference—FFM accelerates enterprise AI adoption and transformation.

AFS: One-Stop No-Code Inference & Optimization Service
1.Locally Optimized FFM Series:
AFS leverages TASC’s proprietary Formosa Foundation Models, trained on supercomputers, to enhance domain-specific understanding and generation capabilities.

2.Integrated Cloud Training & Inference Platform:
Enables custom language model services directly through the platform, tailored to industry needs and local cultural context via high-performance computing.

3.No-Code Fine-Tuning & Deployment:
AFS allows users to fine-tune and deploy customized LLMs without writing code, significantly lowering the adoption barrier.

4.Flexible Deployment Options:
Supports both cloud-based and on-premises deployment, allowing clients to meet their data privacy and security requirements.

TASC: Asia’s First Commercial AIHPC Supercomputer Empowering Industries

TASC is Asia’s first commercial AIHPC (Artificial Intelligence High Performance Computing) supercomputer built to serve industry needs. It offers cross-node AIHPC computing with ultra-high throughput and ultra-low latency via InfiniBand networking, along with a high-efficiency storage architecture. Delivered as an intuitive, on-demand public cloud service, TASC enables users to rapidly establish development environments, significantly reducing R&D time by several folds. By integrating data processing and model optimization, it can deliver up to 100x performance improvement, supporting large-scale, high-volume, and complex workloads across various industrial domains—accelerating model training and field validation for smart applications.

TASC provides a comprehensive ecosystem of AI frameworks, SDKs, and software environments, including ultra-performance container computing (CCS), high-performance computing (HPC), virtual computing (VCS), cloud object storage (COS), and Low-Code/No-Code AI tools. These capabilities accelerate compute-intensive tasks, streamline model deployment, and facilitate collaborative development. Additionally, the platform can integrate with 5G, AIoT, and T-Proof trusted data services to support digital transformation for both government and enterprises.

TASC’s services have earned top-tier cybersecurity ratings and multiple international ISO certifications. They comply with major regulatory frameworks, including the U.S. HIPAA, EU GDPR, and Taiwan’s Classified National Security Information Protection Act. With a 24/7 Security Operations Center (SOC) and enterprise-grade Service Level Agreements (SLAs), TASC ensures sensitive data is securely processed and stored locally, never crossing national borders—delivering exceptional security while supporting procurement via government joint supply contracts.

AI Innovation United (AIU):
Provides startups with opportunities for market positioning validation and brand promotion.
Through the promotion of theAIHPC Accelerator, AIU connects startups with publicly listed and large enterprises to gain deep insights into their needs. Startups propose tailored solutions, which are then shortlisted by the enterprises. Selected startups proceed to present in a live AI Innovation Day Demo Pitch. Ultimately, the most promising startups are chosen to sign MOUs with the enterprises and carry out POC (Proof of Concept) validations.