Imagine being able to put your best sales associate in front of every customer for every interaction. Your best sales associate offers product recommendations and how-to guidance, and explains related products customers might not have considered.
The NVIDIA retail shopping advisor, introduced in this post, is a prebuilt, end-to-end AI workflow. It provides a reference design that demonstrates how to develop a retrieval-augmented generation (RAG) application with large language models (LLMs) that can ingest product catalog data and use some of the latest generative AI features to provide a differentiated experience delivering contextually accurate, human-like answers to customers’ inquiries and recommendation requests.
Retail shopping advisor AI workflow
This retail shopping advisor AI workflow provides enterprises with a fast and advanced way to go from pilot to business value. It includes everything needed to make a consumer shopping experience conversational, precise, and accurate.
The retail shopping advisor reference architecture includes a RAG model that can leverage the most up to date product data when answering customer questions. Also included is a sample dataset of product data from the NVIDIA Employee Gear Store that represents a product catalog. You can use this reference example for adding your own product catalog and related data to create an interactive shopping advisor for your business.
Included with NVIDIA AI Enterprise, NVIDIA NIM microservices ensure rapid, enterprise-grade deployment and optimized model performance. The NVIDIA NeMo Retriever collection of NIM microservices enhance traditional LLM capabilities by effectively using a broad spectrum of enterprise data. This is just a small subset of the suite of software that developers can select from NVIDIA when constructing a shopping advisor application.
NVIDIA NIM is designed to streamline the deployment of generative AI applications, while ensuring security and scalability—key to successful enterprise adoption. NIM encapsulates models and integration code, which are deployed through a Kubernetes Helm chart. These can be deployed on the infrastructure of choice—on-premises or through a Cloud Service Provider. The setup of a NIM provides a path to elevate generative AI applications from a proof-of-concept to production, enabling “zero to inference in just 5 minutes.”
NeMo Retriever contains state-of-the-art, commercially-ready models for retrieval embedding and reranking. Access these microservices through the NVIDIA API catalog. The NVIDIA retail shopping advisor uses a GPU-optimized Milvus Database to store vector embeddings.
With NIM and NeMo Retriever microservices, you can construct your own retail shopping advisor application that can access data in real time, and is optimized to achieve greater recall compared to strictly open source frameworks. The performance benefits of NVIDIA NIM simplify solving a major customer issue of acquiring quality responses to lengthier search queries.
Follow along with a Jupyter Notebook
Included within this workflow is a JupyterLab Notebook server with a sample notebook that shows the solution’s features, so you can quickly prototype and experiment with your own data. Specifically, it covers how to:
- Use LLMs
- Use LLMs with retail product data
- Create embeddings from product information
- Use those embeddings to retrieve products most similar to a given query
- Empower LLMs to make decisions, such as respond normally, or use tools like search or shopping cart APIs
- Collect these pieces into a single ProductAdvisor utility class
- Deploy this in a FastAPI backend and interact with that backend through a chat box in a React application
Get started
Learn more about how to build your own retail shopping advisor. When you’re ready to get started, apply for a 90-day free subscription to access the retail shopping advisor AI workflow for free. To experience additional NVIDIA NIM microservices, visit the NVIDIA API catalog.
For more details about how NVIDIA is helping retailers across novel use cases, check out the on-demand GTC session, How NVIDIA Accelerates Retailers on the Gen AI Journey.
To learn more about building RAG applications at your enterprise, visit NVIDIA/GenerativeAIExamples on GitHub. These examples will help you create your own shopping advisor that can accurately answer domain-specific questions around your enterprise’s products using the most viable information.
