AI-Powered Search

Cohort 1

Nov 2025

March-2026

July 2026

<p>Today’s retrieval systems operate at massive scale (100 billion vectors) and are undergoing exciting AI-driven paradigm shifts to improve retrieval quality, including new approaches for Multivector, Multimodal, Hybrid, and Agentic Search. In this series, we’ll explore both the foundational principles of modern AI Search, as well as what’s being implemented at the frontier of information retrieval.</p>

"Exploring Modern AI Search" Lightning Lesson Series (6-Sessions)

Learn the latest AI Search skills with the authors of "AI-Powered Search"

AI Search Expert | Founder | Author

Trey Grainger

Founder, Searchkernel. Author, "AI-Powered Search". 18+ years leading search.

Principal Search Engineer | Author

Doug Turnbull

Principal Search Engineer, Author of "AI-Powered Search" & "Relevant Search".

AI-Powered Search: Modern Retrieval for Humans & Agents

Staff Software Engineer

Data Scientist

ML Engineer

Generative AI Engineer

Senior Performance Analyst, Instructor

Enterprise Architect 

Data Engineer

Principal NLP Engineer

Cross-encoders: An End-to-end Guide

Radu Gheorghe

Software Engineer, Vespa.ai

Author, AI-Powered Search

As a form of ranking classifier, cross encoders are an essential tool for search relevance ranking.

What are cross-encoders and their uses?

Compared to regular vector search (with bi-encoders), cross-encoders greatly improve semantic ranking, but at a cost.

Trade-offs using cross-encoders vs. standard vector search

Exporting the cross-encoder model (ONNX format), tokenizing doc and query text, & connecting the dots in a rank profile.

An end-to-end walkthrough of using cross-encoders (in Vespa)

Optimization tricks like using multi-stage ranking and running the cross-encoder on a GPU.

Tips and tricks to keep latency and cost in check

Search Trends in 2026 - by the numbers

Dawn Anderson

Founder, Bertey. SEO Consultant.

Co-Author, AI-Powered Search

Detailed analysis of emerging search trends in 2026

How do users actually search in 2026?

Is there a generational gap in AI search? In what domains should you prioritize chat-based search?

Which users use AI search functionality?

Market statistics on what users actually want out of search beyond hope and wishful thinking

Ground product planing in reality

A Bag-of-Documents Model for Query Understanding & Retrieval

Daniel Tunkelang

Independent Search Consultant. LinkedIn / Google / Endeca alum.

Learn an approach to align query and document representations in order to improve retrieval and ranking.

What is the bag-of-documents model?

Build a retrieval model by aggregating queries, results and relevance judgments; and then fine-tuning a base model.

How do you implement the bag-of-documents model?

Learn where the bag-of-documents model outperforms alternative approaches, but also where it can be tripped up.

What are the pros and cons of the bag-of-documents model?

AI Search Fundamentals: Vector Spaces, Matching, & Ranking

Principal ML Engineer in Search

Author, "AI-Powered Search", Founder @ Searchkernel

How does keyword matching work with an inverted index (in contrast with embeddings), and what is a sparse vector space?

Sparse Vector Spaces and Lexical Search

How does search over embeddings work with flat/ANN indexes? What are the many kinds of dense vector spaces that exist?

Dense Vector Spaces and Semantic Search

How do we score how "relevant" documents are for a given query? We'll cover cosine similarity and TF-IDF / BM25 scoring.

Vector Similarity, TF-IDF / BM25, and relevance ranking

We'll summarize the more advanced AI Search methods we'll be teaching in the full AI-Powered Search course

An overview of contemporary and emerging AI Search methods

This session provides all prerequisite knowledge needed for the AI-Powered Search course (please join us!).

You'll be ready for the AI-Powered Search course!

Modern AI Search Techniques You Should Know

Founder @ Searchkernel, Author "AI-Powered Search"

Search and Snuggie Enthusiast

We'll do an overview of key modern AI search techniques to familiarize you with when and how they can improve retrieval.

A survey of modern AI Search techniques

Retrieval is much more than base hybrid search (bm25 + semantic). We'll discuss how to combine other modern techniques.

A Mental Model for Hybrid Search and Retrieval Pipelines

How do modern retrieval systems integrate user behavior and automated relevance feedback?

What is Reflected Intelligence?

Students in the month-long "AI-Powered Search" course learn and implement these (and more). Find out what's in store.

Get a Sneak Peek of the AI-Powered Search Course

Are we all doing "retrieval" wrong?... a candid discussion

Andreas Wagner

Co-Founder & CTO @ searchHub.io

Retrieval use cases often need distinctly different data, algorithmic approaches, and UX per domain and per context.

Agentic and Hybrid Search don’t easily “solve retrieval”

Retrieval is way more than picking great embeddings and keyword search tools. It needs critical additional capabilities.

Search is a dynamic ecosystem

Retrieval engines should constantly adapt to content, domain knowledge, & behavior. They can’t be both static & optimal.

Relevance is a moving target needing ongoing feedback loops

Real strategies, from experimentation & domain-specific evals, to the algorithms that move the needle on search quality.

Learn directly from AI-Powered Search