AI-native platforms and autonomous agents are changing how people discover, decide, and interact with the world. But most AI systems still lack a critical capability: understanding human taste. Galya AI is the taste-based personalization layer for AI-native platforms and personalized agents. We transform how users engage with multimodal content — images, video, text, and audio — into structured preference intelligence that AI systems can understand and act on. At the core of Galya is a VLM-powered multimodal Taste Graph that models relationships between users, content, and aesthetic patterns to infer what people actually like, not just what they search for. Galya is designed as a horizontal infrastructure layer. By exposing our platform through APIs, MCP integrations, Agent Skills, and CLI tools, we enable agents and AI-native platforms to operate with a deeper understanding of their users — allowing them to make decisions that align with individual taste. Rather than replacing existing data infrastructure, Galya sits on top of existing data sources, enriching them with multimodal taste intelligence derived from organic content engagement. As AI agents increasingly act on behalf of users, systems will need a way to compute human taste across content, context, and behavior. Galya is building the infrastructure layer that enables agents to understand preference and see the world through their users’ eyes.