Vectara, a neural network technology platform, announced $20 million in seed funding and also launched its first product for general availability: Neural Search as a Service.
Many existing search engines are keyword-based and attempt to find the best match for the words or phrases used as input. Neural search, however, leverages large language models to understand the intent behind a search that can yield accurate results even if the keywords used are misspelled or inaccurate.
by Vectara Neural Search as a Service is generally available starting today in two editions: Growth, which provides a free tier offering users up to 15,000 queries per month, and Scale, an enterprise-focused tier with a comprehensive set of features that the company claims provides a comprehensive search pipeline capable of running on the biggest websites in the world.
Large stores of unstructured data remain untapped for many organizations, and Vectara says its platform aims to become a premier engine for unlocking insights and value from that data. To achieve this goal, the company has developed zero-hit AI models, or those that enable custom text classifications without requiring annotated or structured training datasets due to the way they leverage networks. of deep learning neurons. Platform enforces deep vector encoding for all indexed content and queries issued, company says blog postenabling a true sense-to-sense response that retrieves relevant information based on a neural understanding of intent and context.
Vectara says it’s launching with a search product because it’s currently the primary way people access information while browsing apps and sites. Some organizations find it difficult to implement search functionality in their applications due to the time and expertise required. “Vectara replaces these outdated approaches with an ML-based composable search pipeline that recognizes searcher intent and radically increases search relevance. Vectara’s ability to innately understand human language means the platform sees through incorrect terminology, acronyms, or misspellings without any predefined language configuration or developer tweaking,” the company said. in a press release.
Vectara is led by CEO Amr Awadallah, an AI scientist, co-founder of Hadoop distributor Cloudera and former vice president of Google. Awadallah is joined by a team of former Google executives and engineers, including co-founders Amin Ahmad and Tallat Shafaat as chief technology officer and chief architect. They are backed by a team experienced in companies such as Cloudera, Elastic, Splunk, MongoDB and Hazelcast.
“We launched Vectara because we saw an opportunity for neural networks to revolutionize the way people find meaning through data,” Awadallah said. “Until recently, however, only the largest organizations with deep ML expertise could access this game-changing technology. With our API-based neural search, developers and engineers can immediately start applying NLP to the challenges they face every day. I have no doubt that the future of research is unequivocally neural research.
Vectara announced that its $20 million seed round was funded by technology and search pioneers from companies such as Slack, Meta and Google, as well as technology venture capital funds including Vertex Ventures. , GTM Capital and Top Harvest Capital. The seed funding will support the launch of Vectara’s first products and the growth of its operations team.
Bradley Horowitz, Investor, Entrepreneur and Vice President of Product at Google, said, “We are at a truly historic moment when the full potential of NLP has reached the level of science fiction. Vectara is the first platform that harnesses that transformative power and makes it practical and useful, giving developers around the world the most advanced AI available. I can’t wait to see the wave of innovation and usefulness that Vectara will unleash.
Are neural networks the next big thing in research?
Algolia acquires Search.io to allow users to “search as they think”
Kyndi Empowers NLP with Search as a Service