Project: Beauty AI Assistant

[sithukaungset]

Project Name

Beauty AI Assistant

Description

An innovative "AI Beauty Assistant" that merges cutting-edge technology with beauty expertise. By integrating the best-selling beauty and skincare products data from Olive Young Korea, complete with product images, I used semantic chunking via Azure Document Intelligence as additional modifications to the data ingestion. This allowed me to break down product information into meaningful sections for better insights. I then vectorized embedding vectors using Azure AI Search, ensuring highly relevant data retrieval. Customer data is securely stored in Azure SQL, where it is used to personalize recommendations and provide tailored skincare advice.

At the heart of this system is a multi-agent architecture. It consists of:

1. Trend Research Agent: Leverages Bing Search to keep up with the latest skincare trends and products.
2. Product Retriever Agent: Uses Azure AI Search's hybrid search and semantic re-ranking to fetch the most relevant product data from our vector storage.
3. Skincare Routine Creator Agent: Synthesizes the data from the other agents to craft personalized skincare routines based on real-time trends and top product recommendations.

I deployed this system using Promptflow and Prompty as the orchestration engine to ensure seamless interactions between agents in Azure AI Studio. With this setup, thousands of beauty enthusiasts can now enjoy tailored beauty advice, discover the hottest products, and get expertly crafted skincare routines—all powered by intelligent agents and LLMs.

The solution integrates an Azure Speech-powered Avatar to deliver an interactive experience, guiding users through product queries and skincare routines for a more intuitive and human-like interface. Additionally, the assistant supports multilingual interactions, automatically detecting the user’s language and responding in either English or Korean, ensuring a seamless experience for a diverse audience.

This dynamic, agentic system offers users the ultimate beauty companion, delivering personalized and data-driven skincare guidance.

Technology & Languages

• JavaScript
• Java
• .NET
• Python
• AI Studio
• AI Search
• PostgreSQL
• Cosmos DB
• Azure SQL

Project Repository URL

https://github.com/sithukaungset/RAG-hack-agent

Deployed Endpoint URL

https://red-pebble-0c942770f.5.azurestaticapps.net/

Project Video

https://www.youtube.com/watch?v=v6ARi5JzJNU

Team Members

sithukaungset

[multispark]

Hello @sithukaungset, thank you for participating in RAG Hack!

The team is working hard to distribute badges. Please have each team member fill out this form:
aka.ms/raghack/badge-dist

Thank you!