Building a Contacts Graph from activity data
summary
Salesforce teams build a contact graph from email and calendar metadata ingested via Kafka and stored in S3, update the raw per-org graph roughly weekly raw-graph updates, and keep edges as collections of timestamped events rather than email bodies. They run Spark jobs for enhanced graphs, apply PageRank-style scoring for influencer identification, serve real-time NLP email insights from streaming, and power features like recommended introductions and contact-mention linking for orgs of up to ~10k users and ~1M daily events.
AI-written summary · about
chapters (7)
- 0:00Introduction to Contacts Graph
- 13:15Platform architecture overview
- 23:59Privacy and data retention trade-offs
- 30:25Recommended connections product
- 33:20Contact mention feature
- 34:44Influencer identification method
- 38:17Key takeaways and Q&A
related talks
21mBuilding a Graph of all US businesses using Spark technologies
33mUnleash the Magic of Salesforce Genie - in Action
30mArtificial Intelligence More Like Artificial Stupidity
15mArchitecting Scalable Data Lineage Graph for Privacy Compliance and Agentic Analysis
56mGraph Analytic Systems
44mHow We Store and Analyze a Large Social Graph
11mReverse ETL with
30mHow Graph Data Science Can Turbocharge Your Knowledge Graph