June 2025 Summaries

Lettria, a leader in document intelligence, achieved a 20-25% accuracy improvement in regulated industries like finance, aerospace, and pharmaceuticals by integrating Qdrant's vector search capabilities with Neo4j's graph-based semantic understanding. Traditional Retrieval-Augmented Generation (RAG) systems fell short in high-stakes environments requiring precise and auditable outputs. Lettria's innovative solution involved building a robust document parsing engine, automatic ontology builder, and a dual ingestion pipeline for vectors and graph enrichment. The system maintained consistent data alignment between Qdrant and Neo4j through a custom transaction mechanism, ensuring atomic updates and conflict resolution in concurrent environments. By flattening payloads and managing over 100 million vectors with low latency, Lettria created a scalable and accurate GraphRAG platform that enhanced explainability and transparency for clients. This approach not only improved performance but also secured high-value contracts by delivering reliable, audit-grade results in complex document intelligence applications.

The Qdrant Vector Data Migration Tool, which has launched in beta, facilitates seamless data migration between different Qdrant instances or from other vector database providers to Qdrant, utilizing live batch streaming for efficiency. This tool allows for straightforward transitions from open-source Qdrant to Qdrant Cloud and across cloud regions, all with a single command, unlike the previous node-specific snapshot method. The migration tool requires matching vector size and distance functions but offers flexibility in modifying collection configurations, such as altering replication factors or quantization methods. It is most effectively operated as a container on a machine with good network connectivity to both the source and target databases, though direct connectivity is unnecessary.

Lawme.ai, a startup at the forefront of LegalTech, has significantly reduced costs and improved performance by transitioning to Qdrant's vector search engine, facilitating the automation of legal workflows with AI assistants. Initially facing challenges with PGVector, Lawme's former database solution, due to performance and compliance issues, the company needed a system that could handle vast data volumes while adhering to strict legal standards. Qdrant offered a solution with features like binary quantization and advanced metadata filtering, which enhanced search speed and accuracy while maintaining compliance with data residency requirements. This switch resulted in a 75% reduction in infrastructure costs and improved query latencies, enabling Lawme to scale efficiently and gain trust from legal clients. With Qdrant's flexible deployment options, Lawme is now positioned for global expansion, maintaining its competitive edge in the legal automation sector through continued innovation and infrastructure optimization.

Qdrant has successfully renewed its SOC 2 Type II certification and achieved HIPAA compliance, reinforcing its commitment to maintaining high standards of security, confidentiality, and compliance for enterprise-scale operations and sensitive data management. The SOC 2 Type II certification, initially obtained in 2024, requires a rigorous 12-month observation of security practices, demonstrating Qdrant's adherence to the American Institute of Certified Public Accountants (AICPA) Trust Services criteria. The new HIPAA certification is crucial for managing Protected Health Information (PHI), ensuring that enterprises using Qdrant's platform can handle sensitive healthcare data in compliance with regulatory standards. These certifications are complemented by Qdrant's suite of security features, including Single Sign-On (SSO), Cloud Role-Based Access Control (RBAC), observability and monitoring tools, and database API key management. Qdrant is committed to ongoing enhancements of its security measures through annual SOC 2 Type II audits and continuous adherence to HIPAA standards, adapting to new threats and industry demands.

LegalTech applications require precise search capabilities due to the complex nature of legal documents and high regulatory demands, which traditional keyword searches often fail to meet. Qdrant offers a vector search solution designed to enhance precision and efficiency in LegalTech environments by integrating features like Filterable Hierarchical Navigable Small World (HNSW) indexing and hybrid search, which combines exact and semantic queries. These tools address the need for speed and accuracy, crucial for high-stakes legal applications, by minimizing unnecessary comparisons and allowing for detailed, token-level similarity estimation. Qdrant also supports scalability and cost-effectiveness through GPU acceleration and vector quantization, enabling LegalTech developers to manage large datasets without compromising performance. The flexibility in deployment options ensures that LegalTech products can meet engineering and compliance requirements while providing enterprise-grade features such as role-based access control and comprehensive monitoring. By leveraging Qdrant's capabilities, LegalTech teams can build scalable and compliant AI applications that effectively balance accuracy, cost, and performance.

ConvoSearch, an AI-powered recommendation engine designed for direct-to-consumer (D2C) e-commerce brands, significantly enhanced its performance and customer revenue by transitioning to Qdrant for vector search operations. Initially hindered by the high latency and limited customizability of Pinecone, ConvoSearch experienced transformative improvements with Qdrant's low-latency queries, extensive metadata storage, and advanced customization capabilities, reducing query latency from 50–100ms to approximately 10ms. This shift allowed ConvoSearch to leverage NVIDIA GPUs for efficient resource optimization and speed, enabling deep product understanding and real-time personalization. Consequently, customers like The Closet Lover and Uncle Reco reported substantial revenue increases, with median revenue growth of 23–24% and a 60% revenue uplift for The Closet Lover. The move to Qdrant not only resolved existing infrastructure challenges but also solidified ConvoSearch's role as a crucial partner for its clients by delivering immediate and tangible business outcomes.

n8n users can now benefit from the official Qdrant node, which simplifies the integration of Qdrant's semantic search capabilities into their workflows by eliminating the need for HTTP request nodes. This development is particularly advantageous for building robust systems such as RAG systems, agentic pipelines, and advanced data analysis tools. The node supports all of Qdrant's features, including hybrid queries, reranking with multivectors, and sophisticated filtering, as provided in the Qdrant 1.14.0 release. Available for both cloud and self-hosted n8n instances from version 1.94.0, the node is easy to install and enhances the ability to perform complex hybrid searches, especially useful in specialized fields like legal and medical domains. A demo video is available to guide users in fusing dense and lexical hybrid search results using Reciprocal Rank Fusion, and the community is encouraged to share feedback and participate in discussions through GitHub and Discord.

Qdrant, in collaboration with Alexey Grigorev and DataTalks.Club, is offering a free, hands-on, 10-week online course focused on building applications using Large Language Models (LLMs). The course will provide practical experience with LLMs, Retrieval-Augmented Generation (RAG), vector search, evaluation, and monitoring. Participants will learn how to create AI systems capable of answering questions about their knowledge base using LLMs and RAG, with the first week covering the fundamentals and the second week delving into vector search. The course, led by Qdrant experts like Evgeniya Sukhodolskaya and Kacper Łukawski, will cover both foundational and advanced concepts including semantic similarity search, hybrid search, multi-vector search, and reranking techniques, with all materials freely accessible online.