Making Money with Data Science
summary
Company develops a customer-facing feature that reduces alert noise by using supervised classification and implicit feedback rather than rules, prototypes a time-based grouping UI, and tests it on PagerDuty-scale data: about 315 million incidents, 589 million notifications, and 1.7 million users. Labeled examples are scarce (28,000 manual merges <0.01% of events), so the team relies on behavioral signals, measures precision/recall and latency, and trades model complexity for explainability and operational SLAs.
AI-written summary · about
chapters (7)
- 0:00Introduction to alarm fatigue
- 3:06Finding customer pain points
- 7:43Vision for data science solution
- 11:13Machine learning fundamentals
- 16:46Prototyping and failure
- 22:24Measuring success and explainability
- 25:54Key takeaways for profitability
related talks
34mTest Driven Alerting: Now with More Data Science!
01h 30mBecome a data-driven Organization with Machine Learning
36mML: Predictive analytics in Ad Engineering
21mData Symphony: Orchestrating Insights in Fintech
22mReal-world applications of machine learning: Beyond the hype
11mData science: Next-gen performance analytics
24mMachine Learning for Analytics Team
23mBig data and big insights