Welcome to Marqo 🙌

Your Unified Platform for Embedding Generation and Search
Marqo is an open-source platform designed to streamline embedding generation, search, and ranking, delivering state-of-the-art retrieval across multimodal data. Also available on Marqo Cloud for seamless integration and scalability into any application or service.

Explore Our Repositories

• Marqo: The core embedding generation and search engine.
• GCL: Generalized Contrastive Learning for multimodal retrieval.
• FashionCLIP: SOTA image and text embeddings for fashion search, recommendations and classification.
• Ecommerce embeddings: SOTA image and text embeddings for ecommerce search, recommendations and classification.
• Ecommerce search application: SOTA multimodal search and recommendations application for ecommerce.

Our Latest Innovations

• Ecommerce product embeddings: Our latest models for ecommerce retrieval.
• Embedding Fine-Tuning with MarqTune: Tailor embeddings to your domain for superior search results.
• FashionSigLIP: Our latest model for fashion retrieval, combining cutting-edge techniques for enhanced search.
• Generalized Contrastive Learning (GCL): A framework for training robust embedding models for multimodal search. Read the GCL paper on arXiv.
• Understanding Recall in HNSW: Insights into understanding and optimizing recall when using hierarchical navigable small worlds (HNSW) in vector search. Read the HNSW paper on arXiv.
• Designing Interfaces for Multimodal Vector Search Applications:. Explore novel UI/UX concepts for implementing multi-modal vector search image search applications. Read the paper on arXiv.
• Marqo-GS-10M multimodal dataset:. Open source multimodal product retrieval dataset of 10M query-product pairs with both text and images. Explore the dataset!