Keynote: Building for AI at Scale: Infra, Integrity, & Innovation
summary
Glean connects Slack Jira Google Docs Salesforce and other silos into a single index with user-aware permission enforcement so answers map back to source provenance and IT can sandbox sensitive content. An agent platform runs autonomous workflows (weekly metrics, standardized interview feedback) and uses LLM judges to auto-evaluate answers while telemetry watches token use and latency. Deployments are self-tuning and self-healing per customer to minimize ops. Cost levers include prompt caching batch inference and selective corpus targeting to avoid unnecessary GPU spend.
AI-written summary · about
chapters (11)
- 0:00AI at scale introduction
- 1:52Glean's enterprise AI assistant
- 7:09Enterprise AI infrastructure
- 10:24Data integrity pitfalls
- 13:55Testing AI agents
- 17:07AI aspirations vs reality
- 20:01Standardizing AI use cases
- 23:11Privacy and trust in AI
- 26:45AI infrastructure costs
- 29:57AI leadership skills
- 32:04Closing remarks
related talks
36mKeynote: AI, ML, Cloud for ML
13mKeynote: How We Save $900 per Day with Self-Hosted AI: Building Scalable Local LLM Infrastructure
41mKeynote
33mKeynote: Fine-tuning our way towards openness in AI
01h 02mKeynote - Building for the Future: Microsoft 365, Agents, & AI
30mKeynote: Fireside Chat
43mKeynote: Building AI in the Open: How the Cloud Native Community Shapes What Comes Next
01h 11mKeynote - Securing AI: Building Trust in the Era of AI