Engineering AI Applications: A Conversation with AI Leaders
summary
Panelists describe running AI in production for healthcare, coding platforms, and finance: 250 enterprise health systems, a coding agent that runs long-lived sandboxes decoupled via Temporal, and jobs that spike to 6 billion tokens in 10 minutes. Key engineering moves are decoupling agent harness from sandbox lifecycles, building in-house ASR to beat off‑the‑shelf accuracy, using a Maximizer routing/prioritization layer for models/providers, and switching from RAG to chunk‑and‑evaluate retrieval when LLM cost/quality allows.
AI-written summary · about
chapters (9)
- 0:00Introduction to AI leaders
- 1:11Building AI systems in production
- 4:22AI agent evolution over 12 months
- 9:20Critical architectural decisions
- 13:30Autonomy vs control in AI agents
- 19:02Missing infrastructure primitives
- 21:48Model selection and optimization
- 26:23Handling availability challenges
- 28:29Future unlocks in AI capabilities
related talks
59mPanel Discussion: Strategies for Scaling AI in Quality Engineering
58mAI in Reality Fireside Chat: Enterprise AI & Open‑Source Innovation
44mPannel Discussion | Agentic Automation: Are AI Agents The Future of Digital Work
34mAccountability in the Age of AI Agents
01h 01mKeynote Panel: Future of AI agents
27mEngineering in the Age of AI: Making AI Work for Us, Not Replace Us
24mAI Governance for developers: A practical framework for deploying autonomous coding agents safely…
26mAI for enterprise teams