From POC to Production - Infrastructure for Machine Learning
summary
Describes how MercadoLibre extended its Fury microservices platform to support machine-learning workflows by building Fury Data Apps. Covers the gap between experimental ML work (data-scientist environments, ETL, model training, GPU access) and production-serving, and explains the tools and services added for building, experimenting, developing and deploying ML systems at MercadoLibre scale.
AI-written summary · about
chapters (10)
- 0:00ML challenge at Mercado Libre
- 3:20Mercado Libre data ecosystem
- 8:14ML infra POC to prod
- 12:04JupyterLab as a service
- 16:27REST APIs for model deployment
- 20:37Modular inference architecture
- 23:06Hybrid ML platform details
- 26:19Platform adoption in teams
- 30:15Future of the platform
- 32:49Q&A
related talks
41mInfrastructure for ML Applications
27mA Look Inside Applied Machine Learning
29mEvolving the ML Platform Organisation at Netflix: A Case Study
56mDevOps for Machine Learning
19mSustaining ML in the enterprise
29mDemystifying Machine Learning in Production: Reasoning about a Large-Scale ML Platform
21mMastering the DevOps Machine Learning Pipeline
40mBuilding highload ML powered service