Unlocking the power of GenAI
summary
Coinbase describes a deployed GenAI customer‑support stack that moved from single‑turn RAG FAQs to multi‑turn dialogues, proactive issue prediction and autonomous execution of documented business procedures, using an agentic chain of LLM agents. Model choice is automated via routine ablation tests across OpenAI Google AWS and open‑source models; monitoring uses DeepEval and RapidEval plus custom LLM‑judged metrics. A simulation test suite, input/output guardrails and an intelligent load balancer across clouds handle surge scenarios (10x traffic) and keep the system compliant.
AI-written summary · about
chapters (8)
- 0:00GenAI at Coinbase
- 3:41GenAI for support
- 8:40Customer support journey
- 13:20Support chatbot layers
- 20:26Testing and securing chatbot
- 23:37Scaling and responsible AI
- 24:20CBGPT platform
- 27:29Q&A
related talks
45m9 Lessons Learned from Deploying GenAI at Scale
41mGetting real DevOps ROI with your Generative AI
33mOperationalizing GenAI Applications with DevOps Practices
23mScaling GenAI Applications and Agentic AI for Business Competitive Differentiation
29mIndustry generative AI in action
24mScalable GenAI in Production
36mGenAI governance in practice
34mOrganized chaos adopting GenAI at Enterprise