构建云原生自主智能 AI 研究应用：深入探讨 pgvector、Remix 和多模态 LLMs

A legacy computer vision project was modernized into an LLM-powered, cloud-native research assistant using PostgreSQL, Redis Pub/Sub, React, and AWS Fargate. Arpad Kish, a full-spectrum software engineer and DevSecOps practitioner, originally built a three-tier client-server application for a content-based image retrieval system in 2014 as part of his Bachelor's thesis. This monolithic C++ and Node.js application utilized traditional computer vision techniques such as SURF descriptors and CIELAB color space clustering. Recently, he integrated foundational concepts from that project with modern technologies to facilitate its transformation.