Human-Centered AI: What I've Learned From Building Two LLM-Powered Products Thi...
summary
Built two LLM products this year that prioritize rapid value: start with summarization and retrieval for quick wins, collect interaction signals (the chat UI surfaces three next-step recommendations) and feed them back to improve models, and abstract complexity with reusable actions so non-experts get consistent results. Real deployments include a midwife assistant handling 20,000 data points per pregnancy that scales to 200 million patients. Technical choices: few-shot + RAG + fine-tune, migrate from private API to smaller tuned models as usage patterns stabilize.
AI-written summary · about
chapters (9)
- 0:00Welcome
- 1:27From investment analyst to AI enthusiast
- 3:36Human-centered AI with midwives
- 5:45Five lessons from LLM products
- 9:40Human-AI collaboration models
- 11:45Abstracting LLM complexities
- 14:15Quick LLM integration
- 15:41Call for everyday AI evaluations
- 17:47Q&A
related talks
39mLeveraging AI and Emerging Technologies for Competitive Advantage
43mAI Development Lifecycle: Learnings of What Changed with LLMs
27mBuilding AI applications: Successes and Lessons
42mUX+AI: Mastering the UX of AI-Powered Products
33mOpening Keynote: LLMs, between fantasy and reality - the poin...
48mMastering the AI Shift: Adaptive Strategies for Software Evolution
32mBeyond LLMs: Building an AI-Powered Decision System at Scale
01h 01mPanel discussion: Practical and responsible applications of AI in businesses