New eBay platform that uses AI to search for images and enable internal innovation

Krylov has allowed the e-commerce site to rethink internal processes and provide users with new tools.

Artificial intelligence: what business leaders need to know
Brandon Vigliarolo explains what artificial intelligence is and how companies can implement AI.

Many of the largest technology companies such as Google, Facebook and Amazon have recognized the value of creating their own AI platforms for both internal and customer-focused services.

More about artificial intelligence

Facebook’s FBLearner Flow helps the social media site to filter offensive messages, while Uber’s Michelangelo gives users time forecasts for food delivery.

To keep up with the competition, eBay has unveiled its AI platform, Krylov, that the company has provided a wide range of new possibilities, from improved translation services to image search.

In a blog post, eBay’s Sanjeev Katariya, vice president and chief architect of the eBay AI and platforms, and Ashok Ramani, director of product management, computer vision, natural and language processing, the establishment of Krylov and how it has changed things both within eBay and for users of the website.

“With computer vision made possible by eBay’s modern AI platform, technology helps you find items based on the click of your camera or an image. Users can go to the eBay app and take a picture of what they are looking for and within milliseconds, the platform shows items that match the image, “Katariya and Ramani wrote in December.

“The user has not only activated computer vision technology, but has also used a number of advanced AI capabilities, including deep learning, distributed training and inferencing. The computer vision algorithm searches more than half a billion images and the 1, Find 4 billion eBay offers and show you the most relevant listings that are visually comparable. ”

SEE: (Artificial intelligence: a guide for managers (free PDF)) (TechRepublic download)

The attempt to create Krylov, named after the famous mathematician Nicolai Krylov, has evolved over time as eBay began to take stock of all its data and how it innovated, they said.

According to Katariya and Ramani, data scientists on eBay are now using Krylov to conduct thousands of model training experiments per month with a variety of AI usage scenarios, including computer vision, natural language processing, merchandising recommendations, buyer personalization, vendor pricing guidance, risk, trust, shipping estimates and more.

EBay data scientists spent weeks and months building models to test new website features and wasted time, money, and energy on processes that needed to be speeded up and streamlined, they said.

Krylov has enabled eBay scientists to automate model training and use the models as a platform through individualized or general inference in days instead of months.

The platform is now being used to improve eBay’s recommendation system and enable powerful image search functions that have opened up a whole new path for the type of searches users can perform. Shoppers can now easily upload a photo of what they want and search for similar items.

The company says it has a lot of data and Krylov enables eBay to harness the power of their information so that they can evolve to better help their users.

With Krylov, eBay scientists have access to notebooks, Tensorflow, PyTorch and H20 and can train models such as BERT or ResNet on a large scale.

“With Krylov, our AI teams can maximize the power of eBay’s huge amount of data storage, both batch and real-time. If you consider data as the fuel for artificial intelligence and machine learning, Krylov is the advanced vehicle powered by that fuel, “wrote Katariya and Ramani.

They said the AI ​​platform was crucial to eBay’s automatic translation technology, which makes a significant contribution to cross-border trade that accounts for nearly 60% of eBay’s international income.

In an interview, Katariya said that a Unified AI Initiative Core Team was established to lead the initiative, and the group included employees working on the AI ​​platform, as well as others working on hardware, networking, storage and data services.

The platform also received a huge amount of input from every part of eBay that could benefit from Krylov, including departments working on engineering in advertising, computer vision, NLP, risk, trust, and marketing.

“We had so many engineers and scientists throughout the company who needed help creating models and pushing their models into production. We needed a completely closed cycle of lifecycle management of machine learning algorithms that was clear. We had A uniform AI platform is needed to really bring data scientists and engineers, modes and management experiments all together, “Katariya said.

He added that the founding of Krylov brought scientists and engineers and platform builders together and was organized by core teams and fellowship programs for machine learning.

Katariya said that every member of the team worked to teach each other, share code, and build on quality.

“Looking back on how we have made progress, I am extremely proud of the collaboration, transparency, internal open source and how training and education have continued to build a truly powerful platform that is global and eBay scale, “he said.

“It took a while for Krylov to grow up, but the goal was clear to ensure that our engineers and scientists around the world were able to access the right data at the right time, whether it was real or not was time or batch oriented data lakes or data warehouses or transaction data in a programmatic way. ”

Katariya noted the huge amount of data that eBay holds, focusing on 1.4 billion offers from the website, 190 markets and 183 million users.

Krylov is now powerful enough to serve as the backbone of eBay’s advertising, merchandising, recommendation and personalization systems.

Katariya explained that eBay’s computer vision is now built on top of Krylov, allowing it to record images and perform object recognition or object detection. This allows the platform to combine image search with text search to arrive at coherent ranking systems.

“Like every parent, I am very proud of what we have been able to achieve with Krylov. It is a significant advance in building such a platform, because it is a polyglot, you can program in multiple different languages,” Katariya said.

“It’s very unique in the way we built it. It’s holistic. The structure itself is built with diversity and diversification in mind. That has resulted in an amazingly powerful platform.”

Cloud Insights newsletter

Your knowledge base for the latest news on AWS, Microsoft Azure, Google Cloud Platform, Docker, SaaS, IaaS, cloud security, containers, the public cloud, the hybrid cloud, the industry cloud and much more.
Delivered on Mondays

Register today

Also see

tampatra, Getty Images / iStockphoto