Voyage AI

Let's keep the momentum going. 🚀 If you haven't heard, Voyage AI, creators of the state-of-the-art embedding model and rerankers, have joined MongoDB. Yes, we know you have questions. What does this mean for your AI applications? Well, join myself and Frank Liu for a webinar tomorrow where we will take you through: 1️⃣ How quantization works to reduce the memory footprint of embeddings dramatically 2️⃣ How MongoDB Atlas Vector Search integrates automatic quantization to manage millions of vector embeddings efficiently 3️⃣ Real-world metrics for retrieval latency, resource utilization, and accuracy across float32, int8, and binary embeddings 4️⃣ Combining binary quantization with a rescoring step yields near float32-level accuracy with a fraction of the computational overhead 5️⃣ Best practices and tips for balancing speed, cost, and precision—especially at the 1M+ embedding scale essential for RAG, semantic search, and recommendation systems Don't miss this opportunity to level up your vector database operations and advance in AI application development. 📅 Date: February 27, 2025 ⏰ Time: 12 P.M. ET 🔗 Register here: https://lnkd.in/ezer88Qd How are you currently handling large-scale vector operations? Share your challenges below!👇 #MongoDB #VoyageAI #VectorDatabases #Quantization #AIEngineering #RAG #SemanticSearch

We're thrilled to join MongoDB and continue pushing the boundaries of AI! If you're looking to boost performance and optimize costs for your AI applications, don't miss this opportunity. Join us for an insightful webinar with MongoDB experts Richmond Alake and Frank Liu, where you'll learn strategic best practices for scalable AI deployments and explore the cutting-edge capabilities of MongoDB Atlas Vector Search. Mark your calendars for February 27, 2025, at 12 P.M. ET.

We are excited to announce that Voyage AI is officially a part of MongoDB! By combining Voyage AI’s best-in-class embeddings and rerankers with MongoDB’s best-in-class database, we will bring seamless, high-quality AI retrieval to more developers—enabling accurate, reliable AI for mission-critical applications. Voyage AI’s embedding models and rerankers will remain available through www.voyageai.com, AWS Marketplace, and Azure Marketplace. Integrations with MongoDB Atlas will launch later this year. https://lnkd.in/eJFs7SmK

Did you know? Many people don’t realize how easy it is to use Voyage AI embeddings with Pinecone! With the Pinecone Model Gallery, you can create an index tailored for the latest Voyage models in just a few clicks. Search for the model, hit “Create Index”, and Pinecone auto-fills key settings like dimensionality & similarity metric. Seamless setup, optimized performance. Check it out: https://lnkd.in/gBnxZFmx

Join our Head of Applied ML Frank Liu and explore about multimodal RAG pipeline! No more complex pipelines just to extract insights—directly process PDFs, tables, and multimodal data without converting everything to text first!

Ready to supercharge your semantic search with Snowflake Cortex and Voyage AI embeddings? here's a deep dive on how to build high accuracy, high performance semantic search directly in Snowflake using vector embeddings from Voyage AI. You’ll learn: • How embeddings enable semantic-level matching • data pipeline architecture in Snowflake Cortex from ingestion to inference • Practical tips for optimizing query performance and embedding storage If you’re an AI /ML engineer, data engineer, or AI builder looking to integrate advanced search capabilities into your Snowflake workflows, this guide is for you! https://lnkd.in/gs-_Qjqe

Looking to simplify AI and semantic search? Check out this article on how to build accurate and fast semantic search with Voyage AI Embeddings in Snowflake Cortex. You’ll learn how to embed, store, and query unstructured data—no extra infrastructure needed—and unlock domain-aware, multilingual embeddings for RAG workflows. https://lnkd.in/gk7qwdnb

Looking forward to participating in this panel today with Akriti Keswani of Airbyte and Frank Liu of Voyage AI, moderated by the coolest guy on the West Coast, Yujian Tang, OSS4AI. There is also another amazing panel discussion between Aiswarya Sankar, Entelligence.AI, Karan Vaidya, Composio, and Vaibhav Gupta, Boundary (YC W23). Don't miss it! If you are interested in attending, register here - https://lnkd.in/gTCAsbH3 #ai #data #agents #aiagents #LLM #datapreparation Stratola LLC

𝐀 𝐜𝐨𝐮𝐩𝐥𝐞 𝐨𝐟 𝐜𝐨𝐦𝐩𝐚𝐧𝐢𝐞𝐬 𝐚𝐫𝐞 𝐛𝐮𝐢𝐥𝐝𝐢𝐧𝐠 𝐦𝐨𝐝𝐞𝐥𝐬 𝐭𝐡𝐚𝐭 𝐚𝐫𝐞 100𝐗 𝐛𝐞𝐭𝐭𝐞𝐫 𝐭𝐡𝐚𝐧 𝐎𝐩𝐞𝐧𝐀𝐈’𝐬! 😲😲😲 DeepSeek AI is the hot topic right now. Sure, there's skepticism about whether they can train a model with just $5 million. I want to share some updates on embedding models, the service every single RAG application needs to transform raw, unstructured data into vectors. Right here in Silicon Valley, startups are doing the unthinkable: building embedding models that are 100𝐗 𝐜𝐡𝐞𝐚𝐩𝐞𝐫 𝐭𝐡𝐚𝐧 𝐎𝐩𝐞𝐧𝐀𝐈's. What’s their secret? Just 𝐡𝐚𝐫𝐝 𝐰𝐨𝐫𝐤 𝐚𝐧𝐝 𝐞𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐞𝐱𝐜𝐞𝐥𝐥𝐞𝐧𝐜𝐞. Take Voyage AI as an example. They’ve revolutionized embeddings with techniques like Matryoshka learning and quantization-aware training, which allow for adjustable dimensions and precision levels. The result? Lower storage costs without sacrificing performance. On top of that, Voyage AI specializes in domain-specific models tailored for industries like finance, law, coding, and multilingual applications, making their offerings highly practical and effective. Voyage AI isn’t an outlier, either. Startups like Cohere and Jina AI are achieving similarly impressive results, proving that innovation doesn’t need to burn hundreds of millons of dollars. Big companies, like cloud providers, can make a fortune by simply commercializing old, mature products, thanks to their market dominance. Startups, with far less funding, far less visibility, and sometimes even attacks from the giants, have only one path forward: hard work and engineering excellence. This isn’t just about carving out a piece of the pie. It’s about shaping the future of AI. For startups, this fight isn’t just for survival. It’s for humanity’s collective future.

📢 We are thrilled to announce that Voyage embeddings are now fully integrated with LanceDB, making it easier than ever to build and scale AI applications with sota embedding models. With this integration, you can seamlessly set up Voyage embedding functions for your LanceDB tables. This feature abstracts away the complexity of generating text embeddings, allowing LanceDB to automatically synchronize vectors for your data. You can learn more about this setup in LanceDB’s documentation. Our Voyage multimodal embeddings are now also available in LanceDB, significantly simplifying the ETL process for document-rich applications. No need for complex parsers or layout analyzers—just specify voyage-multimodal-3 as your embedding function and send in screenshots of documents. It’s that easy. Check out our sample notebook to see how easy it is to get startedhttps://lnkd.in/ghQdWKKe We’re excited to see what you build with this new integration. Let us know how it’s going—we’d love to hear your feedback and see your applications in action. 🚀