About
Designing Hybrid Search Systems, and the engineer behind it.
Professional background
Laszlo Csontos is a software engineer who works on search and retrieval systems: hybrid pipelines that combine lexical and vector retrieval, embedding model selection and fine-tuning, reranking, and the surrounding evaluation and operations work that keeps those systems healthy in production.
He has spent much of his career in backend and platform engineering, most recently focused on information retrieval and RAG infrastructure. The book distills the architectural patterns and trade-offs he has found useful when helping teams move from pure keyword or pure vector search toward hybrid systems.
Why this book
Most of the existing writing on hybrid search lives in two places: vendor documentation that is understandably biased toward the product selling it, and academic papers that are narrow, fragmented, and not written for engineers shipping systems. I kept running into the same gap in my own work and in conversations with other teams: nobody was treating hybrid search as a system design problem, end to end.
This book is my attempt to fill that gap. It starts from the observation that keyword and vector retrieval fail in complementary ways, builds up architectures that take that observation seriously, and covers models, evaluation, operations, and the domain-specific adaptations that matter for RAG, e-commerce, and enterprise knowledge search.
I have tried to write the resource I wish had existed when I first started on retrieval: opinionated where the evidence supports an opinion, neutral where the trade-offs depend on context, and honest about the cases where hybrid search is not worth the added complexity.
The TensorOpt connection
Outside of writing, I also consult on search optimization and RAG systems through tensoropt.ai. This site is about the book, not the consulting practice; if you need engagement work specifically, that is the place to start.
Contact
For press inquiries, review copies, bulk purchases, or speaking invitations, email hello@hybridsearchbook.com.