Laying the foundation for deep learning
summary
Airbnb describes building a production ML stack: an event pipeline called jitney on Kafka, a Hadoop data warehouse with Hive/Presto/Spark, and an open‑sourced ETL scheduler airflow, plus an experiment reporting framework ERF that auto‑captures metrics. Models move from GBDT to deep neural nets; early DNN experiments report a significant booking conversion gain and a model that predicts likely star ratings for listings. An internal AI research lab begins work on image/feel recognition to feed ranking and personalization.
AI-written summary · about
chapters (5)
- 0:00Start
- 3:26Machine Learning and Deep Learning
- 6:51Data Warehouse and Processing
- 8:12Experiment Reporting Framework
- 10:21Search Ranking
related talks
32mTeam Teardown: Organizing and Supporting ML
22mBefore the Model: How Machine Learning Products Start...
33mDeep Learning at Booking.com
40mMachine Learning
39mOperating Deep Learning Pipelines Anywhere Using Kubeflow
42mDeep learning beyond the learning
17mKeynote: Scaling Deep Learning Models in Production Using Kubernetes
44mDeep Learning in Production