Meet Marqo, an open source vector search engine for artificial intelligence applications

Vector databases are the unsung heroes of the modern AI movement, storing unstructured data such as images, video, and text so that humans and systems can search for uncategorized content. They are particularly important for large language models (LLMs) such as GPT-4 (on which ChatGPT is based), largely because of the ability of databases to provide real-time indexing and search as data is created or updated - important for personalization features, recommendation systems, sentiment analysis, and more.

The growing demand for generative AI has brought a slew of vector database startups into the spotlight and raised huge sums of money along the way. In April alone, Pinecone and Weaviate raised $100 million and $50 million, respectively, to develop their vector databases, and in the same month, vector database startups Chroma and Qdrant received $18 million and $7.5 million, respectively, in seed funding. And late last year, Zilliz, the primary developer of the open source vector database Milvus, raised $60 million in funding.

So it's clear that companies working to help infrastructure keep up with the rapid evolution of AI are in high demand. Australian startup Marqo aims to capitalize on this by offering a more holistic "end-to-end" approach to vector search.

Pain point

Founded in Melbourne last June, Marqo is the brainchild of Jesse Clark, formerly a lead machine learning specialist in Amazon's robotics division in Seattle, and Tom Hammer, formerly a database software engineer at Amazon Web Services (AWS) in Sydney.

The essence of Marqo's mission is to solve the problem of unstructured data, which by some estimates accounts for up to 90% of all data created. As more and more people turn to generative artificial intelligence to answer their online queries or create new images and artwork, it only reinforces the need for new tools to make sense of it all.

Marqo's main advantage over existing competitors is that it promises a complete set of intelligent vector search capabilities out of the box, including vector generation, storage and retrieval. This means that Marqo allows its users to dispense with third-party vector generation tools from companies like OpenAI or Hugging Face and offer everything through a single API.

"Vector search is hard to implement - vector databases are only one piece of the puzzle, and it's difficult for developers to put together all the necessary components to create vector search with optimal relevance, latency, and reliability," explained Marqo co-founder and CEO Tom Hamer in an email to TechCrunch. "Marqo provides a comprehensive system that brings all of these components together, solving a major problem for developers."

Moreover, search engines are only as good as the results they produce, which means that relevance, accuracy and "relevance" are integral to any information storage and retrieval system. And that's what Hammer says Marqo offers from the start.

"If developers want to continuously improve the relevance of search results, they have to manually train new AI models to generate vectors," he continues. Marqo's continuous learning technology will allow it to automatically improve search based on user engagement - such as clicks, "add to cart," etc. This is especially important for e-commerce and other end-user search use cases."

Last year, Marqo raised £660,000 ($840,000) in pre-funding, and today it announced it has raised a new $4.4 million in seed funding to double its commercial efforts. This includes a new cloud service that will be officially unveiled to the public today in addition to Marqo's existing open source project.

The open source factor

Like many of its competitors, Marqo's open source ethos was a very deliberate move, designed to cater to a community of developers who can mess around with the product to see if it's right for them. In turn, this means they are more likely to recommend the product to influencers in their company and even contribute to its development.

"I'm a firm believer that open source product development leads to higher quality," says Hamer, "Building Marqo on open source has allowed us to have close feedback from our users and iterate very quickly to create a product that developers really want. Open source is also a great customer engagement channel. Customers can see exactly what they are buying, try it for free, and see if Marqo is right for their needs.

However, creating an open source product usually requires a lot of resources, both human and infrastructural. And that's where Marqo Cloud comes in.

"Self-hosting an open source product is a great option for users who do not require real-time search and have a small number of end users, or to create a proof-of-concept," Hamer continues. "Marqo's cloud platform provides the infrastructure, maintenance and operation of cloud resources for our customers, ensuring optimal performance and cost efficiency."

Although Marqo is an Australian startup, it has registered its parent company in the UK, where its first investor, Creator Fund, is based. The company also has a small office in London, which currently has one employee, but it is looking to expand to include sales, marketing and customer support to support its ambitions across Europe.