Considerations for maturing AI initiatives from PoC to Production
summary
Distinguishes ML vs deep learning by real numbers: ML uses tens to hundreds of terabytes up to a petabyte and tens-to-thousands of servers for feature models, while deep learning uses multiple petabytes and clusters of hundreds to thousands of GPUs for images, audio and NLP. Recommends consolidating data stores, decouple compute from storage, tier hot/cold data, add lineage/quality governance, run inference at the edge and training in data centers, and adopt containerized DevOps pipelines that move workloads from workstation to warm cluster.
AI-written summary · about
chapters (4)
related talks
41mInfrastructure for ML Applications
25mImportance of Data Maturity before embracing AI
06mMachine learning is about your data and deployment
31mProductionizing ML with ML Ops and Cloud AI
19mSustaining ML in the enterprise
05mAI and ML are changing traditional business
29mDemystifying Machine Learning in Production: Reasoning about a Large-Scale ML Platform
30mChallenges of machine learning: The new ML lifecycle