How Bayer Built an Enterprise-Scale Search Engine with Qdrant - Qdrant

This post details Bayer's journey building an enterprise-scale search engine using Qdrant. Key contributions include: 1. **Scalability and Compliance:** Transitioning from self-hosted Qdrant to Qdrant Hybrid Cloud to meet strict life sciences compliance requirements while managing operational burden for 116,000 employees. 2. **Omnimodal Data Handling:** Implementing a pipeline to ingest and search across diverse data formats (text, images, audio, video) by generating multi-vector embeddings from figures in scientific PDFs and indexing transcribed audio/video. 3. **Agentic Retrieval:** Exposing the Qdrant API directly to agentic applications (LangGraph-based harness) for deep retrieval loops, where each tool call is backed by a Qdrant collection, emphasizing low per-query latency. 4. **Hybrid Search for Grounding:** Utilizing hybrid search (keyword + semantic) as a core component for grounding LLMs and reducing hallucinations, crucial for agentic applications. 5. **Caching for Cost Control:** Implementing caching strategies to manage agent costs by reducing redundant search queries. 6. **Performance and Evolution:** Demonstrating the ability to scale to millions of messages and adapt to evolving data models and AI workloads, highlighting Qdrant's role as a foundational retrieval engine.