State-of-the-art CLIP/SigLIP embedding models finetuned for the fashion domain. +57% increase in evaluation metrics vs FashionCLIP 2.0.

Academic Information Retrieval and Automated Theory Development with Large Language Models

Implemented four retrieval models for search engine implementation and evaluated their performance.

Application to search scientific papers and graphically summarize them using LLM using Information Retrieval concepts

Proposing a novel approach to music recommendation that takes the audio of the user as input, converts it to text and performs sentiment analysis, tf-idf, and normalization. We then use content-based and collaborative filtering techniques to recommend songs

Codebase of Cross-Lingual Neural Databases