Develop AI-powered Applications with OpenAI Embeddings and Azure Search

Go beyond an LLM's knowledge cutoff. Learn to build AI applications that use your proprietary data with OpenAI embeddings and Azure Search.

#1about 3 minutes

Understanding embedding vectors as numerical representations

Embedding vectors convert complex concepts like text or personality into multi-dimensional numerical arrays, enabling comparison and clustering.

#2about 7 minutes

Working with the OpenAI embeddings API and cosine similarity

The OpenAI API provides an endpoint to generate a 1,536-dimensional vector for a given text, and vector similarity can be efficiently calculated using a dot product.

#3about 5 minutes

Building custom applications with the OpenAI chat API

The chat completions API allows developers to build custom applications by sending a model the entire chat history, including system prompts and user messages.

#4about 3 minutes

Implementing the Retrieval-Augmented Generation (RAG) pattern

The RAG pattern enhances LLM responses by first retrieving relevant facts from a private knowledge base using vector search and then injecting that context into the prompt.

#5about 4 minutes

Demo overview of building a school wiki assistant

A practical demonstration shows how to build a Q&A assistant for a school's private wiki using a crawler, an indexer, and a query application.

#6about 8 minutes

Step 1: Crawling and pre-processing the source data

The first step in the RAG pipeline involves building a custom crawler to extract, clean, and convert source data into a usable format like Markdown.

#7about 6 minutes

Step 2: Indexing embeddings into a vector database

An indexer application iterates through pre-processed documents, calculates their embeddings via the OpenAI API, and stores them in Azure Cognitive Search for fast retrieval.

#8about 5 minutes

Step 3: Querying the system using the RAG pattern

The query application generates an embedding for the user's question, performs a vector search to find relevant documents, and injects them into a system prompt for the LLM.

#9about 5 minutes

Live demonstration of the wiki Q&A assistant

The command-line assistant successfully answers specific questions about school policies by retrieving information from the wiki, even handling multi-language queries.

#10about 13 minutes

Q&A on embedding calculation, ethics, and tooling

The speaker answers audience questions about how embeddings are calculated, ensuring answer correctness, responsible AI development, and recommended developer tools.