Building the Payments Foundation for Tomorrow’s AI Economy
summary
Stripe describes a payments stack that already runs Transformer-based risk model in production (migrated in 2022), powers identity checks across ~50 countries, and claims over a third of the company used an internal LLM chat in 3 weeks, with ~60 live LLM apps now. Customer wins include 40% faster payments page for OpenAI. The docs search uses a RAG-powered pipeline with embeddings, hierarchical indexing and custom parsers, deployed after ~a year of iteration and now used by thousands in production.
AI-written summary · about
chapters (8)
- 0:00Introduction to Stripe
- 1:12Challenges in scaling payments
- 2:55AI and ML in Stripe’s products
- 4:18Internal LLM experimentation
- 5:20LLMs in user-facing products
- 6:23Documentation with RAG
- 8:50Building accurate RAG systems
- 10:36Future of AI in payments
related talks
15mGrowing
45mStories from a quarter-century in payments
50mThe Future of On-chain Payments
21mCollision 2015 - John Collison of Stripe with Alexandra Suich of The Economist
30mSecureAI: A Scalable, Secure, and Compliant AI Solution
16mMy Bank Meets AI: Smarter Decisions Beyond Fancier Spreadsheets
14mKeynote: Building AI with Open Source & Hugging Face
17mLaying the foundation for deep learning