Project-ELK

Inspiration

"Businesses may want another way to analyze whether a new product will be successful." This system can be used by Square sellers to improve their businesses in a number of ways. For example, sellers can use the system to: • Set more competitive prices • Determine which products to stock in their inventory • market their products to the right customers • Identify new product opportunities

What it does

Upload a product photo and get sales insights for similar products. Square sellers can now upload a photo of their new product and get information about how similar products have sold in the past. This information can be helpful for sellers when setting prices, determining inventory levels, and marketing their new products.

How we built it

The system scrapes e-commerce sites to gather items and inserts them into Square's test sandbox environment. This allows sellers to test their products in a safe environment before they go live.
The system uses a Selenium web scraper to get images of all items. Selenium is a web automation tool that can be used to control a web browser and interact with web pages.
The system uses Generative AI annotation to generate keywords for item images. Generative AI is a type of artificial intelligence that can be used to create new content, such as text, images, and code.
The system stores the keywords in a Google Cloud database. Google Cloud is a cloud computing platform that offers a variety of services, including databases.
The system uses the Gecko model in Vertex AI to generate text embeddings for each keyword and store them in the database. Vertex AI is a managed machine learning platform from Google Cloud. The Gecko model is a text embedding model that can be used to represent text data as vectors of numbers.
When a user uploads a new image, the system repeats steps 3-5. The system then compares the text embeddings for the new item with the embeddings in the database to find similar items.
The system gets the historical sales data for similar items.
The system generates metrics and returns them to the user. The metrics may include average sales price, total number of units sold, best-selling colours and sizes, sales trends over time, and popular related products.