Data science at The New York Times
summary
The New York Times applies machine learning end to end: Python and TensorFlow for training, Go for real‑time ranking, BigQuery and Google Cloud Storage for data, Redis/memorystore for vectors. Models include LDA topic model, ALS collaborative filtering and a linear contextual bandit; infrastructure serves recommendations to apps, web and two algorithmic emails. A Slack tool called Blossom guides editors. Built systems increase engagement by multiple tens of percent, monetize advertising products and regularly pay for several data scientists’ salaries, and they work in production.
AI-written summary · about
chapters (8)
- 0:00Data science at NYT
- 3:53Data scientist role
- 10:09Machine learning for predictions and prescriptions
- 13:42NYT future plans
- 14:29Customer journey personalization
- 25:10Product vs feasibility
- 29:59Recommendation algorithms
- 37:45Cross-functional teams
related talks
38mDeploying Data Science for Distribution of The New York Times
29mSeveral People Are Tuning: Data and Machine Learning at
01h 17mOne workshop that data scientists don't want you to attend...
38mData Science from the Trenches
01h 30mBecome a data-driven Organization with Machine Learning
19mData science for real problems
22mReal-world applications of machine learning: Beyond the hype
40mCausal machine learning for a smart paywall at The New York Times