1
CTO/Co-Founder RevelAI Health we are transforming how patient care is delivered through AI. Happy to chat about building AI products that actually solve real problems, or share war stories about HIPAA compliance adventures. Always up for brainstorming
1. Vector space embedding + n clusters -> sample 10 questions per cluster for LLM summarization. Works well when themes/intents aren't predefined.
2. Using LLM to predict question intent/labels. More expensive but better when we have defined themes.
Which approach would you recommend?