ML Pipelines
summary
An overview of machine-learning pipelines focusing on current problems and missing capabilities, plus which emerging technologies could help. The talk examines concrete pipeline examples and discusses implications for Java ecosystems.
AI-written summary · about
chapters (24)
- 0:00ML pipelines introduction
- 1:24Defining ML pipeline
- 3:10Pipeline stability issues
- 4:14Simple pipeline example
- 6:20ML pipeline maturity levels
- 8:05Pipeline tools
- 10:16Orchestrator at Odnoklassniki
- 11:45Spark in pipelines
- 13:11Airflow as standard
- 14:59Converting pipelines to online
- 16:03Streaming issues
- 17:57Feature Store and data
- 19:44Current Feature Store solutions
- 21:06Scalability problems
- 23:10ML team roles
- 26:21T-shaped specialists
- 28:50Skills and tools
- 31:00Simplicity vs responsibility
- 34:16Pipeline quality control
- 37:03Accountability for results
- 38:27Implementing standards
- 40:55Pipeline ownership
- 42:41Practice prevalence
- 44:27Recommendations for small business
related talks
44mIndustrial ML – Overview of the Technologies Available for Machine Learning
49mMachine Learning Pipelines: What No One Tells You About Building AI
56mML Pipeline: Structuring Data Analysis Projects
33mBuilding Resilient (ML) Pipelines for MLOps
47mML in Java, YES it's possible!
55mOrganizing the Machine Learning Lifecycle with MLflow
43mML application lifecycle
17mCrafting Superior ML Pipelines with SapientML