Five lessons Etsy learned building on Microsoft AI
summary
Etsy describes using the Azure OpenAI service and GPT models to build Guided Search and Gift Mode, solving query rewriting, title rewriting and a mission clarity classifier. Prompt engineering techniques (few‑shot, chain‑of‑thought, system messages) raise mission classification from ~45% to ~87% and JSON constraining plus fine tuning improves output validity from ~90% to 98%+. A PTU plus pay‑as‑you‑go mix cuts real search latency from ~5s down to ~2s while keeping monthly costs predictable.
AI-written summary · about
chapters (5)
- 0:00AI at Etsy and lessons
- 18:36Tools and solutions
- 25:03Optimizing for impact
- 34:32Engineering constraints
- 39:36Future steps and Q&A
related talks
29mTransformation in a New Era of AI
01h 01mUnlocking the Potential of AI
37mLeveraging Generative AI for Enhanced E-commerce
16mSolve for AI
48mGetting started with generative AI using Azure OpenAI Service
27mBuilding AI applications: Successes and Lessons
01h 12mChatGPT and OpenAI API solutions: 👑 My Favourite Implementations
01h 10mSolutions prototyping using Azure AI Studio