Keynote: AI, ML, Cloud for ML
summary
Argues that AI must live in every layer of a large search stack — indexing, deduplication, attribute extraction, query understanding, retrieval, ranking and UI — because small gains compound into huge business impact; cites 170 use cases collected across the stack. Describes concrete constraints and deployments: hundreds of online classifiers, hybrid batch/stream/compiled model runtimes, typeahead using suffix arrays, conversational interfaces tied to ASR, and strict performance targets like 20 ms latency, 20k QPS. Explains team roles, SRE needs and platform tradeoffs.
AI-written summary · about
chapters (16)
- 0:00Start
- 3:59Typical Search Engine - High Level View
- 7:43Al is not separable from engineering
- 14:15Indexing - attributes extraction
- 14:55Index - statistical tasks
- 15:06Index - data quality
- 16:13Query understanding Classification
- 17:05Query understanding: entity and attribute extraction
- 18:52Al price prediction
- 19:57Building Search Teams
- 20:45Do you need a search team?
- 21:56Typical roles in the Search team
- 22:59Skills - search infrastructure
- 25:49Skills - Operations / SRE
- 27:41Skills - Product managers
- 29:53Running a search team
related talks
27mKeynote - Google Cloud, Machine Learning
26mKeynote: AI in the Real World: Today and Tomorrow
34mKeynote: Building for AI at Scale: Infra, Integrity, & Innovation
39mKeynote Panel State of AI in the Enterprise
25mKeynote Kickoff
12mDigital & Data/AI Strategy and Transformation
23mThe State of GenAI & ML in the Cloud Native Ecosystem
12mKeynote: Building Enterprise AI/ML Platform on Kubernetes