LandingAI is building the Visual AI platform—an agentic AI framework that empowers users to solve complex visual tasks through natural language prompts and media inputs like images and videos. The system autonomously selects and orchestrates vision models, generating deployable code to automate future workflows. A key challenge in this development is the need for substantial computational power to fine-tune agents and run reinforcement learning loops for continuous performance gains, all while maintaining rapid iteration speeds to keep pace with innovation.
To meet these technical demands, LandingAI partnered with FPT AI Factory, leveraging its high-performance Metal Cloud powered by NVIDIA H100 GPUs. With strong SLA guarantees, robust infrastructure, and expert support from FPT, LandingAI successfully conducted supervised fine-tuning and reinforcement learning at scale. The H100 infrastructure integrated seamlessly into existing pipelines, allowing the team to iterate quickly on model architectures and agent behavior. FPT’s competitive pricing also enabled effective cost management without compromising on performance.
With the efficiency of FPT’s H100-powered infrastructure, LandingAI has trained several state-of-the-art internal models for the Visual AI platform. The enhanced computing capability significantly improved model accuracy and responsiveness across diverse visual tasks. This partnership has accelerated the deployment of customer-facing features and expanded the generalization capabilities of LandingAI’s visual models, all while maintaining high service stability and availability. The collaboration brings LandingAI’s VisionAgent one step closer to scalable real-world deployment.
Cookie | Duration | Description |
---|---|---|
cookielawinfo-checbox-analytics | 11 months | |
cookielawinfo-checbox-functional | 11 months | |
cookielawinfo-checbox-others | 11 months | |
cookielawinfo-checkbox-necessary | 11 months | |
cookielawinfo-checkbox-performance | 11 months | |
viewed_cookie_policy | 11 months |