​About Event

​Generic embedding models get you "kind of similar." Real product search needs the right brand, size, color, and variant. That gap is where fine-tuned retrieval models earn their keep.

​In this hands-on workshop, you'll build a product search lab around a custom fine-tuned retrieval model trained on realistic e-commerce data. You'll see where the baseline embedding model breaks, what fine-tuning fixes, and how to measure the lift with real relevance metrics. You'll also see how the fine-tuned model fits alongside sparse and hybrid search in a production retrieval stack.

​We'll dig into the failure modes generic models can't fix: semantically close but commercially wrong results, missed attributes, weak exact-match behavior, irrelevant substitutes, and bad ranking. You'll learn how hard negatives and domain-specific training data drive most of the quality gains, and how to tell whether a fine-tuned model is actually pulling its weight in production. The same model also powers "similar products" and preference-based recommendations, so we'll cover that path too.

​No pre-work required. Every participant gets a ready-to-use VM with the environment, data, models, and tools already installed.

​​AGENDA

​5:00 – 5:40 PM | Arrival, Registration & Networking
Check-in, refreshments, and access to the demo environment

​5:40 – 6:00 PM | Introduction

​6:00 – 6:20 PM | Antoine Balliet, Senior Engineer - Gorgias
An agent for an agent: automating the setup of an AI support agent.

​6:20 – 7:20 PM | Dylan Couzon, DevRel Engineer - Qdrant
Fine-Tuning AI Search for E-commerce

​7:20 – 7:50 PM | Ramesh Lekshmynarayanan, Founder - Green Catapult
Commercially Wrong: Retrieval Failures in Agentic AI Systems

​7:50 – 8:30 PM | Q&A

​8:30 – 9:00 PM | Wrap-Up & Networking
Key takeaways, Q&A, and networking

​What You’ll Learn

• ​How a fine-tuned retrieval model outperforms generic embeddings on real product data

• ​How hard negatives and domain-specific examples drive the quality gains

• ​How to evaluate a fine-tuned model with practical relevance metrics

• ​How to diagnose retrieval failures and tell "similar" apart from "relevant"

• ​Where sparse, dense, and hybrid search fit alongside the fine-tuned model

• ​How the same model supports recommendations, RAG, and agentic workflows

​​WHO SHOULD ATTEND

​Builders with hands-on experience in RAG, vector search, semantic search, or retrieval pipelines who want to go deeper on fine-tuning and search quality.

​Seats are limited