On Vector Search and Generative AI. Q&A with Arvind Jain
“ I believe that a sense of clarity and alignment on shared objectives is vital to workplace happiness. People are happier and more fulfilled when they’re able to contribute effectively. That’s why our mission at Glean is to empower individuals by connecting them with the information and people they need to thrive at work. “
Q1. You co-founded, Glean, a company to make it easy for people to find the information they need to be more productive and happier at work. Can you please elaborate what do you mean with “more productive and happier at work.”?
Almost ten years ago, I left Google to build Rubrik, which quickly became one of the fastest growing companies in cloud data management. As we scaled up, I was frustrated because although our headcount was growing, productivity was not increasing at a commensurate pace. I dug in to try to identify the problem, and conducted a series of workplace surveys. It turned out that our employees were constantly struggling to find the information, tools and people they needed to do their best work.
It wasn’t a new problem — we built the world’s best search engine for consumers at Google, but we didn’t have anything nearly as effective to help us find answers at work. I’m not a mad scientist, but I’m obsessed with productivity — that’s why I founded Glean.
SaaS applications have revolutionized knowledge work, but their proliferation within the enterprise has also unintentionally created knowledge silos that are almost impossible to organize and navigate. Last year, Forrester put out a report that showed that the top reason employees feel disengaged from work is that data and/or information is hard to find. In 2021, U.S. employee engagement dropped for the first time in a decade according to Gallup. In the new hybrid/remote reality, employee experience is suffering. People are struggling to find answers to their questions, to access the information they need, and stay connected not only to company knowledge, but also to one another.
I believe that a sense of clarity and alignment on shared objectives is vital to workplace happiness. People are happier and more fulfilled when they’re able to contribute effectively. That’s why our mission at Glean is to empower individuals by connecting them with the information and people they need to thrive at work.
Q2. What does it mean in practice to perform a Vector search powered by deep learning-based LLMs?
Performing a Vector search powered by deep learning-based LLMs (Large Language Models) refers to a search method that leverages numerical representations of text, called embeddings, to better understand the semantic information and relationships between concepts. This approach enables a smarter and more accurate search experience than traditional keyword-matching methods.
Glean’s unique differentiator is that we create advanced embedding models for enterprises by training them on each company’s specific knowledge base. These enterprise-specific embedding models are combined with traditional information retrieval methods and advanced personalization to create a hybrid system that delivers truly unmatched enterprise search capabilities for our customers.
Q3. How does it help with the semantic understanding for natural language queries
The fine-tuning process trains on natural language (among other data), so the resulting models are very well-suited to do semantic understanding when the search queries are made in natural language.
Q4. How is Glean supposed to be trained?
Glean trains our core models in-house. Each customer gets their own dedicated Glean instance, which automatically adapts itself to their specific language domain.
Q5. No need for manual fine-tuning. How do you then know if the results are accurate and relevant?
One of Glean’s key differentiators is that we provide full referenceability — meaning that with each result or answer we deliver to our users, we also provide links to the source information across documents, conversations and applications to ensure trust and confidence.
Additionally, we have the proof from our customers’ feedback and metrics — once a company connects all of its data to Glean, we see a high degree of satisfied queries for user searches.
Q6. One of your company tag line is ” Search with generative AI you can trust.”. How do ensure trust for Generative AI?
Generative AI has the potential to be a powerful tool for knowledge workers, but general-purpose models aren’t built with the enterprise in mind. Accuracy, security and referenceability are critical in business, so to be genuinely useful, generative AI needs to understand not only the content, but also the context, the relationships among people, a company’s internal language, as well as privacy and security parameters.
We’ve spent four years developing technology to meet these requirements. Our trusted knowledge model is built around three essential pillars:
- Company knowledge and context: we retrain deep learning language models on a company’s unique knowledge base and develop a thorough understanding of content, internal language, people and the relationships within an organization. This enables us to recognize nuances like how people collaborate, how each piece of information relates to another, and what information is most relevant to each user.
- Permissions and data governance: we take into account all real-time enterprise data permissions and governance rules and ensure that users only have access to information that they’re allowed to see.
- Full referenceability: for generative AI results, we provide full linking to source information across documents, conversations and applications.
Q7. How do you position your company in the market space of Generative AI – Enterprise AI Search?
Glean believes search is the key to unlocking the value of Generative AI for enterprises. Over the past four years, we’ve built a powerful, unified search technology which spans SaaS apps. We retrain deep learning language models on each company’s specific knowledge base, and in this way, develop a deep understanding of context, lexicon, behavior, and relationships with others that’s uniquely tuned to your workplace and adheres to your data governance policies: a trusted knowledge model.
Glean has built the only technology that solves for the accuracy, data security and privacy concerns that have plagued Chat GPT and Generative AI in the enterprise.
Glean is easy to set up, and does not require expensive professional services contracts or custom development effort. Every company wants to figure out their generative AI strategy, and Glean is ready to help.
Arvind Jain is the CEO of Glean, an AI-powered workplace search company he co-founded to make it easy for people to find the information they need to be more productive and happier at work. Prior to Glean, Arvind co-founded and led R&D at Rubrik, one the fastest growing companies in cloud data management. Arvind also spent over a decade at Google as a distinguished engineer, where he led teams in Google’s Search, Maps, and YouTube products. Earlier in his career, Arvind held leadership positions at Akamai and Microsoft. He earned his BTech in Computer Science from the Indian Institute of Technology, Delhi, and his Masters in Computer Science from the University of Washington.