Workforce AI Gets a Hardware Upgrade: How YY Group Is Using NVIDIA Blackwell to Train Smarter Hiring Models

AI-powered workforce management is moving beyond cloud APIs — and one Asia-based company is proving that proprietary hardware can be a serious competitive edge.

YY Group Holding (NASDAQ: YYGH), an AI-native workforce and integrated facility management (IFM) platform operating across Asia, has announced a strategic investment in local high-performance computing infrastructure built on NVIDIA GeForce RTX 5090 GPUs and the NVIDIA Blackwell architecture.

The investment is designed to power internal development and fine-tuning of YY Group’s proprietary, vertical-specific AI models — the kind that can’t be built effectively on generic off-the-shelf APIs.

What the Hardware Does

CEO Mike Fu explained the reasoning simply: “True workforce AI requires more than generic automation wrapper APIs.” The new infrastructure gives YY Group’s engineering teams the processing speed to:

• Fine-tune 7B–14B parameter open-weight models using QLoRA and NVIDIA CUDA optimization, tailored to regional labor market nuances across Asia
• Accelerate candidate-to-job matching using NVIDIA TensorRT for faster semantic search and reranking with ultra-low latency
• Run high-throughput local serving via vLLM and NVIDIA-accelerated APIs that plug into YY Group’s existing product suite

Key Takeaways

• Proprietary models beat generic AI for complex, domain-specific workflows like workforce optimization and IFM operations.
• On-premise hardware investment signals a shift from cloud-only AI strategies toward localized, faster, and more controllable model development.
• Asia’s workforce market is a key focus — YY Group operates across hospitality, F&B, retail, and service sectors where rapid and accurate staffing is critical.

For companies building vertical AI strategies, this is a case study worth watching. The question of “buy vs. build” in AI is shifting — and hardware ownership is increasingly part of the answer.

🔗 Read the full article on Manila Times