#PowerToTheDeveloper Creating Meaningful AI Assistants and CRV’s Investment in Voyage AI
By Murat Bicer and Brian Zhan
When you have a companion who keeps rambling on and on about the same topics over and over again, it wears on you and erodes your trust in that person. The stakes are just as high, and the feelings are just as valid, in the enterprise AI realm.
A customer chatbot that keeps regurgitating the same useless information while ignoring your actual question is infuriating. A legal AI assistant is essentially worthless if it can’t reference relevant legal cases. Coding chatbots that are untethered from a company’s code base become meaningless without solid grounding. Businesses leveraging AI without teeth come across as amateurs to their sophisticated customer bases. As consumers become increasingly pressed for time their tolerance for untethered AI tools that lack an area of expertise wanes. Thankfully we’re entering an era where useful two way dialogue with tech can become a reality.
To be effective in enterprise settings, AI models often require access to specialized knowledge. Traditionally developers have addressed this by adding relevant information directly to the prompt, however, as the volume of necessary background information grows a more scalable approach becomes important. This is where Retrieval-Augmented Generation (RAG) comes into play. RAG is a technique that enhances an AI model’s responses by retrieving pertinent information from a knowledge base and incorporating it into the user’s prompt. Voyage AI uses state of the art vector embeddings and rerankers, pre-trained on domain-specific data, with the option to custom train on the company’s data as well, to provide the best RAG system.
Given the pressing need, and overwhelming excitement in this space, we’re thrilled to announce that CRV just led Voyage AI’s $20 million Series A.
Co-founders and executives at leading AI companies including Harvey and Replit rave about Voyage AI’s embedding models and rerankers. Check out Harvey’s recent partnership blog post here.
On top of that, executives at the leading AI platforms like Anthropic, Snowflake and Databricks have chosen to promote Voyage AI’s embedding models and rerankers to their customers.
Voyage AI’s embedding models are the best available. Voyage AI’s latest embedding model, voyage-3, outperforms OpenAI’s vector embedding, OpenAI-v3-large, by 7.55 percent on average across all evaluated domains shown below with 3x lower vectorDB costs. The model performance improves when using domain-specific models, which are extensively trained on data from their respective domains. This includes models voyage-code-2, voyage-law-2, voyage-finance-2 and voyage-multilingual-2. See the below comparisons on Voyage AI’s model performance and cost. You can find more information on the methodology here.
Tengyu Ma, the co-founder of Voyage AI, is someone we’ve been eagerly paying close attention to for a long time. He pioneered the field of deep learning as a Princeton Ph.D., was a visiting scientist at Meta AI and also an assistant professor at Stanford. We’ve been equally impressed by the team of leading AI researchers and programmers he’s recruited to Voyage AI.
The AI revolution is just beginning and with Voyage AI enterprises will have the tools they need to navigate this new frontier safely, efficiently and with unprecedented insights. As we stand on the cusp of a new era in enterprise AI, Voyage AI is poised to play a pivotal role in shaping how businesses harness the power of large language models. We at CRV are excited to partner with Tengyu and his team on this journey to make enterprise ready AI a reality.
If you’re a founder building in enterprise tech, AI, cybersecurity, SaaS or any other innovative sector, we’d love to connect with you. Our team is passionate about supporting visionary entrepreneurs who are solving complex challenges and pushing the boundaries of what’s possible.
