Vespa Training

Vespa Fundamentals

Agenda

Introduction to Vespa

• What Vespa is, when to use it, and when not to.
• Overview of Vespa’s architecture, query model, and deployment model.
• A first look at Vespa’s API and tools.

Lexical Search

• How attributes and indexes differ and when to use each.
• Designing rank profiles for different use cases.
• Hands-on exploration with Vespa’s YQL query language.

Working with Tensors

• Introduction to tensors as Vespa’s core data structure.
• How to index and query tensors for similarity search and recommendations.
• Understanding tensor operations for ranking and personalization.

Hybrid Search

• Combining lexical and vector search for retrieval-augmented generation.
• Using embedders to create embeddings on the fly.
• Ranking strategies for hybrid use cases.

Grouping and Aggregation

• Using grouping to explore, filter, and diversify results.
• Aggregating metrics for analytics and monitoring.

Performance and Scaling 101

• Scaling Vespa to billions of documents.
• Designing for low-latency queries and high throughput.
• Best practices for tuning performance.

Vespa for E-Commerce Ranking

Agenda

E-Commerce Search Fundamentals

• Combining BM25, nativeRank, document fields, and user preferences in a two-phase rank profile.
• Designing for relevance, business goals, and customer experience.

Vector Search in Practice

• Choosing and fine-tuning embedding models for product catalogs.
• Evaluating semantic search quality using golden sets with LLMs as judges.

Learning to Rank

• Training and deploying gradient boosted decision tree (GBDT) models.
• Using multi-phase ranking to incorporate multiple signals.

Real-Time Recommendations

• Implementing recommendations with sparse tensors, co-occurrence analysis, and basket similarity.
• Scaling recommendation workloads with Vespa’s query pipeline.

Chunking for RAG

• Leveraging Vespa’s built-in chunking for product descriptions, reviews, and rich content.
• Combining document-level and chunk-level scoring for higher-quality retrieval.

Advanced Hybrid Search

• Accelerating semantic search with Matryoshka embeddings and binarization.
• Using float vectors as re-rankers for improved relevance.

Model Tuning

• Why off-the-shelf encoders often underperform in retail search.
• Techniques for fine-tuning encoders for product and user intent.

Late Interaction Models

• Introducing ColPali for visual retrieval directly on product images, without OCR pipelines.
• Cross-Encoders and Multi-Phase Ranking
• Deploying ONNX cross-encoders as third-phase re-rankers.
• Running GPU-accelerated inference in Vespa for maximum precision.

The RAG Blueprint

A Guided Best Practice for Building Production-Ready Retrieval-Augmented Generation systems.

While Vespa’s training courses provide hands-on instruction, The RAG Blueprint serves as a self-guided framework for teams designing large-scale retrieval-augmented generation applications.

The RAG Blueprint covers every stage of system design and deployment, from chunking and hybrid retrieval to phased ranking, embedding strategies, and performance tuning. It draws on proven patterns used in production by leading AI-driven companies and is intended to help engineers, architects, and technical managers shorten the path from prototype to production.

Key benefits of The RAG Blueprint

• Step-by-step guidance on architecting robust RAG systems.
• Practical examples of hybrid search, late interaction models, and on-the-fly embeddings.
• Insights on scaling to billions of documents while keeping latency low.
• Recommendations for evaluation, monitoring, and continuous improvement.

Who it’s for: Teams exploring or scaling RAG, whether for chatbots, agentic AI workflows, or domain-specific discovery applications.