In data management we use indexes in searching, to retrieve the answers in faster than linear complexity. A representative of such index could be the B-tree, which is used for classical attribute data, or M-tree/M-index from the similarity searching domain handling high-dimensional complex descriptors. Recently, in their paper: "Learned Metric Index (LMI) - Proposition of learned indexing for unstructured data", a research team at FI MU argued for the benefits of replacing these traditional rigid indexes with an index composed of a series of machine learning models, built for complex data. A practical example of a complex data type where LMI can be applied is protein sequences and structures. Considering that the recent breakthroughs in protein structure solving introduced mainly by AlphaFold 2 generate a substantial amount of data, we can assume that a system capable of an efficient and effective organization, searching, and visual presentation is very valuable for the researchers in this field. This thesis aims to develop a web application that would allow for browsing protein structures by their similarity. It will support specifying a protein by its name, finding and displaying similar proteins, and providing interactive 3D visualizations of the proteins using existing libraries. The end product will be accessible at a publicly available URL, to benefit the broader research community. The application will be created with the help of widely used frameworks and technologies - any of React, Angular, Blazor, Flask, or Node.js, based on the student's preferences, and deployed within the Kubernetes ecosystem in cooperation with CERIT-SC.