Source: Hebbia

Guide:

• Faced with the challenge of professionals in finance, law and other industries to deal with massive and complex information, Hebbia uses its original Matrix AI technology to break through the limitations of traditional search and provide in-depth and transparent solutions. It not only goes beyond keyword matching, but also disassembles complex problems into clear steps through multi-step proxy execution, so as to achieve true intelligent analysis and decision support. Hebbia's interface allows users to work with AI to optimize search results while ensuring the security and privacy of data processing.
  
• Hebbia's multi-modal processing capabilities, integrating multiple data formats such as PDF, images, emails, and slides, ensure comprehensiveness and accuracy of information. At the same time, seamless integration with popular tools such as Dropbox, Google, Microsoft, etc., greatly improves work efficiency. Hebbia is not only a search tool, but also a disruptive knowledge partner.
  
• Since its founding in 2020 by a Stanford dropout Ph.D. student, Hebbia AI has rapidly grown into a high-profile AI company. It raised $1.1 million in Pre-Seed funding in October 2020, and in June 2022, the company closed another $30 million Series A funding round, attracting participation from several top venture capital firms and well-known angel investors. In June 2024, Hebbia successfully raised $130 million in Series B funding led by a16z, and the company's valuation soared to nearly $800 million. Hebbia AI is rapidly commercialized, and its paying customers now include the world's top private equity firms, hedge funds, legal firms, pharmaceutical companies, and government departments.
  

Source: Hebbia

01 Hebbia – Automate workflows, break through the limitations of traditional search, and provide transparent and efficient knowledge work solutions for professional fields

"As a lawyer by profession, I spent thousands of hours in front of dual monitor screens. My right screen is used for "real" work, such as drafting legal opinions or pleadings for clients, which is highly professional. My left screen is used to assist me in my work, such as flipping through hundreds of pages of due diligence reports, searching through a huge library of relevant cases, and combing through thousands of evidentiary materials with the aim of uncovering only a few key legal issues and evidentiary clues. This is neither challenging nor pleasant, but it is necessary at this stage. ”

Professionals such as financial analysts, lawyers, and consultants are faced with the challenge of dealing with massive amounts of complex information on a daily basis, and while AI is being integrated into these industries to optimize workflows, such as easy-to-build RAG chatbots and numerous information search and document processing applications on the market, most AI products still face challenges such as insufficient unstructured data processing, accuracy (hallucination) issues, and privacy protection when dealing with complex problems that require in-depth analysis. These limitations also expose the limitations of some AI products in terms of professionalism and understanding of deep-seated issues.

Source: Hebbia

Hebbia is an AI-driven enterprise search platform that provides solutions for knowledge workersFounded in 2020 in New York, United States. Its original information processing technology, Matrix AI, breaks through the limitations of traditional language models, and is no longer limited by the limitations of single file processing and search results, providing ready-to-use solutions for professionals such as finance and law. Unlike other search tools, it is able to provide answers in the form of citations after entering the search information, all in the form of a spreadsheet. For each document (row!) ), and you'll get a set of questions (columns!). In addition to summarizing each query, it also shows the source and individual steps that led to the conclusion, which are built in a completely transparent way and allow users to verify, provide transparency into the AI processing process, build user trust and confidence in the output, and allow users to mine and solve problems efficiently and confidently, regardless of the amount of data.

Hebbia, who had deployed the first operational RAG (Retrieval Enhanced Generation) system in 2020, found that RAG and any chat-based AI failed for 84% of real user queries. When dealing with real-world problems, the current limitations of document search products focus on:

1. The questions asked by the user are too complex and require deep analytical skills that are not answered in the existing documentation, they should be completely new insights generated by the AI itself.

2. The search process is too complicated and requires many steps (not just one search) to get a result that can be used directly.

3. The answer is not just text, but charts, tables, and images, and structured data cannot be directly output.

4. The output is very important, and the fault tolerance rate is very low, so the process cannot rely on the black box subsystem.

02 Reinvent the future of search and knowledge work with intelligent solutions that go beyond keyword search

According to Hebbia, AGI is a product problem, not a technical problem, and it does not depend on the continuous increase in the number of model parameters, but rather is designed as a product problem that can process large amounts of information, automate workflows, and present the results of your work the way you want.

Hebbia AI's technology differentiates itself by breaking down complex problems into understandable steps, and both the AI and the user can clearly navigate the solution.

Major search engines such as Google, Bing, and Baidu rely on keyword matching to sort search results by analyzing vocabulary, user data, and semantic maps in web pages. The limitation of this model is that the user needs to have some prior knowledge of the information being sought.

Hebbia's AI technology can go beyond keyword restrictions and break down user problems into multi-step proxy executionProvide in-depth answers. For example, for the question "What is the meaning of life?" Google searches for related blogs, and Hebbia digs deeper and provides academic resources for famous existentialist philosophers. Or a task like "Please evaluate this investment opportunity" is not a search prompt, but rather an analytical process. According to the company, Hebbia's search performance is 57% higher than the performance of the state-of-the-art search engine algorithms.

Source: X

Hebbia AI's product differentiator: Communicate your thought process and limitations to users instead of trusting them.

For industries such as finance and law, which require high specialization and have a very low tolerance for error, even the most powerful models are used, and users cannot trust the results generated by the models without understanding their thought processes.

Matrix mimics a human workflow to manipulate the retrieved data, including summarizing, analyzing, abstracting, comparing, and synthesizing. These steps are performed by independent AI agents, each responsible for a specific step.

In the Matrix interface, users can work with AI agents to modify the original data (source) and processing steps (steps) to optimize the final search or analysis results. At the same time, Hebbia is also able to provide a detailed explanation of each step taken by the AI in the decision-making and execution process, so that users can clearly understand how the AI makes specific decisions.

Source: Hebbia

Matrix AI is multimodal and has the flexibility to process data in multiple formats, including PDFs, images, emails, and slides. It enables flexible data processing and selection between full-text LLMs and vision models, and the system is able to detect, parse, and locate information, and select the most appropriate model for different tasks to achieve optimal processing results.

Hebbia offers integrations with popular software tools like Dropbox, Google, and Microsoft to enhance user experience and streamline workflows that enable seamless data transfer and increase productivity.

The company argues that search engines like Google have access to only a small fraction of the world's data, about 4 percent, and the remaining 96 percent needs to be obtained by partnering with data vendors, who are often reluctant to fully disclose the valuable data they have. Hebbia's algorithm doesn't rely on fast access to keywords in the text, which means it can process data without revealing the content of the data, and it can encrypt and store everything in its index, ensuring the security of the data.

03 Hebbia is becoming a key force in the transformation of the financial and legal industries

Traditional search giants such as Google and Baidu have an extremely profitable business model, thanks to their leadership in the search market, and can use keyword search and page rankings to effectively sell high-value ads, but for users, they cannot maximize the benefits.

As a To B software provider, Hebbia's business model is mainly focused on specific areas, such as finance, law, and healthcare. This kind of vertical market is actually very suitable for AI to enter, first of all, the data is highly privatized, the degree of professionalism is high, the moat is deep, and the current ability of AI can basically handle tedious basic work, and the value of the customer's single unit is also very high.

Over the past 18 months, Hebbia AI has grown revenue 15x, increased headcount by 5x, driven more than 2% of OpenAI's daily trading volume, and laid the groundwork for clients to redefine the way they work. The world's leading private equity, hedge funds, legal, pharmaceutical companies, and defense governments, such as Centerview Partners, Oak Hill Advisors, Fisher Phillips, American Industrial Partners, and US Air Force, are already paying to use Hebbia AI to create value for their companies.

Clients of asset managers and investment banks say that analysis that used to take 2-3 hours now takes only 2-3 minutes and can produce new outputs that they could not have imagined before. It was mentioned that he would be afraid to remove Hebbia because his team would be upset and could face attrition.

Another client shared, "Hebbia has changed the way we do business, far exceeding our expectations. Working with Hebbia is like another member of our team. Customer feedback has been overwhelmingly positive. ”

04 A talented boy dropped out of school to start a business and was supported by top venture capital

Hebbia's founder, George Sivulka, began his first job at NASA at the age of 16 working on satellite software for mine detection, followed by physics research on dark matter detectors for the United States Department of Energy at Stanford University, where he became the fastest undergraduate graduate to earn a mathematics degree. He then dropped out of Stanford University for a Ph.D. in electrical engineering to found Hebbia.

Source: LinkedIn

Tim Lupo, co-founder and CEO, graduated from the University of Southern California in United States.

Source: LinkedIn

Lukas Schmit, co-founder and CTO, graduated from the University of Southern California, United States.

Source: LinkedIn

Hebbia raised $1.1 million in Pre-Seed funding in October 2020, led by Floodgate. In June 2022, it raised $30 million in Series A funding, led by Index Ventures, with participation from Radical Ventures and Abstract Ventures, as well as a group of well-known individual investors, including Yahoo founder Jerry Yang, one of Google's first investors, Ram Shriram, a well-known United States hedge fund investor, Marty Chavez, CFO of Goldman Sachs, and Kevin Warsh, former executive secretary of the White House National Economic Council. In June 2024, Hebbia completed its latest Series B funding round, with $130 million invested by a16z, Index Ventures, Google Ventres, and Peter Thiel, valuing the company at nearly $800 million.

"We are entering a new SaaS era, moving from Software-as-a-Service to Service-as-a-Software. So far, software has helped knowledge workers do their jobs better ......But in the not-too-distant future, an AI agent will do all the work for them entirely. Whether it's reviewing more deals, automating quarterly earnings reviews, or leveraging cross-company knowledge to drive faster customer value, AI colleagues will soon become an integral part of our day-to-day work. We are excited to lead Hebbia's latest round of funding." I am thrilled to join their board and support the company as we unlock the potential of LLMs in financial services and knowledge work." So said Alex Immerman, partner of the growth team at a16z, the newest lead investor.

This article is from Xinzhi self-media and does not represent the views and positions of Business Xinzhi.If there is any suspicion of infringement, please contact the administrator of the Business News Platform.Contact: system@shangyexinzhi.com