In this episode of The Stack Overflow Podcast, host Ryan Donovan interviews Mahir Yavuz, the Senior Director of Engineering at Etsy, about how Etsy uses machine learning to manage and categorize its vast inventory of unique, non-SKU products. Mahir discusses the challenges of dealing with a marketplace of millions of sellers and hundreds of millions of listings that change daily. He explains Etsy's approach to enriching product metadata using GenAI and LLMs, the various guardrails and evaluation techniques they employ to ensure accuracy, and the combination of traditional ML processes with cutting-edge AI. The conversation also covers Etsy's search and recommendation systems, the importance of intent in guiding user experience, and the use of AI-powered curation to introduce surprise and diversity in product discovery. Finally, Mahir touches on how Etsy is exploring the use of generative AI to improve internal productivity and create agentic commerce experiences for both buyers and sellers.
Sign in to continue reading, translating and more.
Continue