Summary
In the latest episode of Dr. Pawd, Mike Fisher, the CTO of Etsy, discussed the company's AI-powered e-commerce model and how they quickly adapted to become a top supplier of masks during the pandemic. Etsy's mission is to keep commerce human by bringing together buyers and sellers and finding the right product for each buyer's journey. With about 100 million active buyers and 7.5 million sellers, Etsy has a massive amount of items, including handmade, unique, and vintage products.
Etsy is a big data machine learning company that processes over six billion events of data a day to power recommendations, search, and other parts of the buyer journey. In 2017, Mike Fisher was brought in by CEO Josh Silverman to analyze and balance infrastructure and product development investment for the company. The company migrated all its supporting infrastructure, processes, big data, and machine learning to the cloud in under two years, which allowed more engineers to work on higher-level tasks and focus on improving the customer experience.
To personalize the user experience, the company uses advanced AI systems in its search engine such as neural information retrieval (NIR) and x-walk. The company also uses a semantic candidate set and a learning-to-rank algorithm to reduce candidate selection and re-rank items to get the most likely to be purchased at the top. Etsy has been introducing real-time features for model inference to allow in-session personalization, rather than just relying on historical data.
To train their models, Etsy uses both positive and negative signatures from analyzing items that were high in the list but didn't get clicked or favored. In addition to text, Etsy is heavily focusing on computer vision to identify image content, style, and color of listed items, as well as flagging mature content. The use of computer vision for style and image recognition has enabled recommendations based on real-time clicks and interests.
In addition to current AR features, Etsy is looking to expand further into using computer vision for image recognition. The team at Etsy plans to use AI and machine learning on almost every part of the site, from the start of the seller's journey to the buyer's side, including accurate product description, visually enhanced product images, correct tags, quality of photos, categories, project ideas, search, recommendations, and gifting.
Overall, Etsy's mission is to keep commerce human and to provide special, customized, and personalized gifts to customers. The company's infrastructure team starts preparing for the holiday season in July, working on high availability and performance and monitoring everything. Etsy has a blameless culture where the team works together to solve any problems by uncovering the root cause of the issue, learning from it, and taking steps to prevent it from happening again. Mike Fisher's background in the army brings lessons of leadership, staying calm, and breaking down complicated situations to simple parts to his work at Etsy, ensuring the company can navigate any challenges that arise.
