Trey Grainger & Doug Turnbull

Cohort 1

Nov 2025

March-2026

July 2026

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.

"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

ML Engineer

Generative AI Engineer

Instructor

Enterprise Architect

Data Engineer

Principal NLP Engineer

Development Lead

Software Engineer i11

Director of Engineering

Lucidworks

Presearch

Optimized AI Conference

CareerBuilder

Searchkernel

Shopify.com

OpenSource Connections

Reddit

Wikipedia

Yelp

AI Search is among of the most in-demand AI Skills. AI systems need RAG, and implementing great retrieval can often have a bigger impact than prompt engineering for improving your AI and Agentic systems. Most search systems barely scratch the surface of what's possible. While others struggle with irrelevant results and frustrated users, you'll master the cutting-edge techniques that power the world's most successful search experiences—from Reddit and Spotify to the Fortune 500.Why This Course ExistsThe Problem: Many traditional search systems are stuck in the 2010s. But modern AI courses hyper focus on text embeddings and ignore many of the fundamentals of information retrieval. The result? Engineers who can't bridge the gap between amazing demos and production systems that actually work.Our Solution: Stop downloading random embeddings from Hugging Face or OpenAI and hoping for the best. Learn to systematically analyze your users' actual intent, integrate domain context, and build search that truly understands what people (and agents) need—not just what they type. This course helps you build cutting-edge AI Search systems that actually work.Ready to level-up your AI Search skills?

Level up your Agentic Search & Retrieval skills with leading AI Search experts

Optimize your Agentic Search and Retrieval Augmented Generation (RAG) systems

Go beyond the query to understand intent

Fill in holes in your modern information retrieval knowledge

Improve relevance with crowdsourced ranking models built from click-stream data

Optimize AI-Powered Search Models for Production

Build ranking features & automate judgements with click models & Active Learning

Examples will be taught in Python, it's a good idea to have basic Python familiarity under your belt

Basic Python Programming

For those new to search/retrieval, we provide an optional intro session covering core matching & ranking concepts (BM25, vector spaces, etc)

(Included Complimentary with the Course): Search Fundamentals

AI Engineers optimizing search ranking to measurably improve user search satisfaction

Search Practitioners that want to learn to systematically improve search relevance

Technology leaders hoping to plan search experiments to maximize clicks and conversions

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.

How the experts actually implement retrieval

Agentic Search: What's the big deal?

How users can go to tediously scanning search results, to having a agentic partner research on their behalf

What problem agentic search solves

Agents can reason about simple search tools (ie a keyword index). They may not a complex search backend

Why you may not need that big search stack

Emerging patterns in agentic search from Recursive Language Models, giving agents their own coding abilities, and more

How agents can write their own search code

Optimizing Search & Data Processing with Self-hosted SLMs

Daniel Svonava

CEO at Superlinked

Founder at Searchkernel, Author "AI-Powered Search".

Principal Search Consultant

See what tasks are ideal for switching from LLMs to SLMs with the same or better quality.

Search & Data Processing with Small Language Models

Lessons from wrapping SGLang, vLLM, TensorRT and why we chose to rewrite most of the OSS model code.

Architecting an inference engine

How to support 35+ model architectures without going crazy and how to run LoRAs in production.

Support 100s of task-specific models

Design an auto-scaled multi-model multi-modal cluster to drive all GenAI projects in your business.

1 million tokens per second in an OSS K8s cluster

Searching 100 billion vectors in object storage

Nathan VanBenschoten

Chief Architect at turbopuffer

Principal Search Engineer

Learn how turbopuffer was built from first principles as a fast, yet stateless query tier on object storage like S3.

Architecting a modern search database from first principles

Many engines over-rely on expensive memory or SSDs. Learn where the true bottlenecks lie and how to optimize for scale.

The critical compute vs. bandwidth tradeoffs and bottlenecks

See how multi-level centroid-based indexing overcomes typical bandwidth bottlenecks on slower object storage tiers.

How hierarchical clustering overcomes bandwidth bottlenecks

Discover how RaBitQ compresses vectors 16-32x while preserving recall to power fast search over 100 billion vectors.

Learn directly from Trey Grainger & Doug Turnbull