Rexera Leverages the Managed Vector Database Zilliz Cloud to Power Its AI Agents, Accelerating the Real Estate Closing Process

In the United States, real estate transactions are often complex and document-heavy, involving multiple stakeholders and lengthy closing cycles.Tech startup Rexera believes AI can simplify these workflows. By leveraging Amazon Bedrock and Zilliz Cloud vector database, Rexera built an advanced AI agent system that automates critical steps—such as retrieving HOA documents, processing mortgage payoffs, and conducting lien searches.The solution significantly reduces time and costs, and is now used by over 350 real estate firms, supporting about 1% of all residential transactions in the U.S.

Real estate transactions involve large volumes of complex text, including legal documents that require careful handling to ensure accuracy and protect the interests of all parties. To streamline this process, tech startup Rexera offers an AI-driven solution that addresses key challenges in property transactions—such as processing legal documents, extracting key information, and identifying potential risks. Rexera’s AI helps real estate professionals reduce manual errors and improve transaction accuracy, making AI a vital assistant in the closing process.

From day one, Rexera built its platform on Amazon Web Services (AWS), using Amazon Bedrock to host various large language models that power its AI agents. As adoption grew, so did the data volume—Rexera’s agents now handle over 10,000 tasks daily and process millions of pages monthly, supporting real-time real estate transactions across the U.S.

With rapid growth came new challenges. Initially, Rexera stored vector embeddings using Deep Lake on Amazon S3, but this setup could no longer meet performance demands. The team tried managing Milvus independently but found the operational overhead and scaling limitations costly, especially during peak transaction periods. Rexera needed a more efficient, scalable way to manage vector data—one that could seamlessly integrate with its existing Amazon Bedrock-based architecture.
 

To address growing performance demands, Rexera evaluated several vector database solutions—including Weaviate and Chroma—and ultimately selected Zilliz Cloud, an AWS Partner solution built on Milvus. This replaced their previous use of Elasticsearch and Deep Lake to support document intelligence. Rexera chose Zilliz Cloud for its high performance, scalability, and seamless integration with the existing AWS environment. The solution was easily deployed through AWS Marketplace and now supports hybrid search capabilities for multiple models hosted on Amazon Bedrock.

“Hybrid search gave us a 40% accuracy lift. We eliminated Elasticsearch. And we don’t think twice about scaling. Zilliz Cloud made all of that possible.”
——Sasidhar Janaki, Senior Software Engineer, Rexera

With support from the Zilliz team, Rexera migrated its vector storage and search infrastructure. In the real estate AI agent workflow, Rexera uses Amazon Bedrock models (via OpenRouter) to encode documents into vector embeddings, which are then stored centrally in Zilliz Cloud. For orchestration, Rexera relies on LangChain tools and LangGraph workflows to combine and coordinate specialized agents. To deliver fast and relevant results, Rexera utilizes Zilliz Cloud’s hybrid search, which combines vector similarity, full-text search, and structured metadata filtering—enabling retrieval of the most relevant content from thousands of pages in milliseconds.

For critical transactions, Rexera also integrates cross-model validation, using multiple foundation models—such as Claude and Llama—on Amazon Bedrock to ensure result accuracy and offer multi-perspective insights. This enhances compliance and risk mitigation for real estate professionals.

As documents evolve over a transaction’s lifecycle, updated embeddings continuously flow into Zilliz Cloud, ensuring search results stay current and continue to support both the models on Amazon Bedrock and Rexera’s AI agents.
 

By migrating from Elasticsearch and Deep Lake to managed services from AWS and Zilliz Cloud, Rexera significantly streamlined its infrastructure. With Zilliz Cloud as the unified vector database, Rexera now powers high-performance search for models running on Amazon Bedrock—enabling its AI agents to become efficient assistants for real estate professionals. This architecture simplification, combined with managed services, led to a 50% reduction in overall costs, making Rexera’s AI solution more competitive and cost-effective for industry clients.

In production, the combined solution of Amazon Bedrock and Zilliz Cloud reduced latency by 30% compared to previously evaluated alternatives. Even with more than 10,000 daily tasks and millions of document pages processed monthly, Rexera’s AI agents consistently deliver reliable, timely results across U.S. real estate transactions.

By integrating Amazon Bedrock’s embedding capabilities with Zilliz Cloud’s hybrid search, Rexera also achieved a 40% improvement in retrieval accuracy, allowing for more precise identification of critical steps in real estate workflows. Today, Rexera’s AI agents help automate up to 99% of the transaction workflow, enabling faster document review, minimizing human errors, and reducing both time and labor costs—ultimately driving transaction efficiency across the industry.

Looking ahead, Rexera plans to deepen its strategic collaboration with AWS and Zilliz. With Amazon Bedrock, Rexera will continue optimizing its use of embedding and foundation models to enhance agent performance. Meanwhile, Zilliz Cloud’s multi-tenant architecture and strict data isolation will support scalable, customized AI solutions for more clients. Together, the three parties will expand the AI agent capabilities to include advanced search, dynamic summarization, compliance prediction, and more—helping real estate professionals unlock long-term value from AI-driven innovation.

Rexera is transforming the real estate industry by leveraging AI agents to automate error-prone manual settlement tasks and complex processes in property transactions. By minimizing human intervention, Rexera enables real estate professionals to accelerate deal cycles, reduce friction, and lower operational costs—bringing greater speed and efficiency to the entire transaction process.

Zilliz is a global leader in vector database and AI technologies, and the creator of the open-source vector database Milvus (GitHub Stars > 35K). Zilliz provides open-source and fully managed enterprise-level solutions, aiming to support next-generation AI applications. Headquartered in San Francisco, Zilliz has raised $112 million in funding from top investors such as Prosperity7 Ventures, Pavilion Capital, and Hillhouse Capital.

Organizations of all sizes across all industries are transforming their businesses and delivering on their missions every day using AWS. Contact our experts and start your own AWS journey today.