welcome

My name is Tyler Suard, and I work at a Fortune 500 company that’s more than a hundred years old with 50,000 employees around the globe. Over the course of a year and a half, I built a retrieval augmented generation (RAG) chatbot for them—discovering, in the process, that for every single method that worked, four others fell flat. Simply put, there wasn’t a tried-and-true handbook on building an enterprise RAG system… until now.

After plenty of trial and error, we ended up with a chatbot that searches 50 million records from 12 different databases, plus 100,000 PDF pages, yet still manages to respond in 10–30 seconds. Thousands of people use it daily, and we get maybe one complaint every two weeks—and it’s usually the user’s fault! When Manning reached out saying other companies struggle to build RAG systems, I jumped at the chance to share the specifics: from organizing massive datasets to reducing LLM hallucinations, to pulling off top-notch answers in real-world conditions.

I’m joined on this project by Darshil Modi, a brilliant AI engineer who holds a Master’s in Computer Science (AI specialization) from Santa Clara University. Darshil fills in many of the gaps in my knowledge, offering insights on advanced vector database usage and security concerns. Together, we’ll show you how to: