Global 
United States 
Multimodal Search and Retrieval with GenAI
Category:
GenAI / FMCG
The Problem Statement
An FMCG giant that extends its services across the globe and is widely known for its extensive portfolio of trusted brands and products required the development of a multimodal data search and retrieval solution.
The Solution
With the increase of data collection in different modalities (Audio, Videos, Image and Text), search and retrieval of data across the modalities is the need of the hour. KritiKal developed a solution that employs a Multi-modality Large Embedding Model to encode audio transcript, image captions, image tags, and videos of user interactions. The innovative system utilized these latent representations to assist seamless searching across different modalities as well as improving the accuracy and relevance of product suggestions. The application ingests these embeddings into database and facilitates a multimodal search and retrieval across different data modalities through query based on any modality of data. A web-based user interface helps in visualizing retrieved results and also provides interactive tools like word cloud, etc.
Benefits Delivered
The multimodal data search and retrieval solution effectively encodes various data types, enhances search and recommends the right skin care product to the consumers by comparing real-time images and analysis results with relevant innovative products present in the database as and when applicable in the case.
Technology Used
- Multimodal LLMs
- DALL-E and BERT