Previous research has established theoretical foundations and experimental validations of personalised similarity search using Mahalanobis distance metric models. However, a full-fledged implementation demonstrating practicality and scalability in real-world scenarios is yet to be developed.

The objectives of the thesis:

• Implementing an end-to-end personalised similarity search system based on the existing theoretical and experimental results.

• Developing a robust architecture capable of efficiently handling personalized queries using learned Mahalanobis matrices while utilizing a common Euclidean-based indexing structure.

• Evaluating the system implementation on criteria such as scalability, response time, retrieval accuracy, and robustness under real-world usage scenarios.