Sustainable Machine Learning by Carl Zeiss AG
summary
Carl Zeiss builds and runs ML systems as software: central Accelerator team with project managers, data scientists and MLE/Ops, aiming to grow to ~100 people. Production stack standardizes on Azure Machine Learning Studio, Azure DevOps, Terraform and Docker; experiments move from notebooks into tested Python packages and Docker images. Key products: Visufit (9 cameras, ~45M face points, 5,500 3D frames, live in ~30 countries), microscopy pipelines that can produce ~4.6 TB/day and a predictive service running on 3000 microscopes across 60+ countries.
AI-written summary · about
chapters (7)
- 0:00Sustainable ML at Carl Zeiss
- 7:24ML development challenges
- 15:37Visofit recommendation engine
- 18:08Microscopy image analysis
- 24:58Predictive maintenance for microscopes
- 32:44ML applications summary
- 33:27Q&A on ML production
related talks
32mThe ML system endeavors
30mCultivating a company-wide machine learning culture
27mTransforming pharmaceutical R&D through industrialized machine learning
27mHow to make ML teams 80% more productive.
30mMachine learning for computer vision in pharmaceutical manufacturing
21mML Infrastructure for Autonomous Vehicles
43mA lap around Azure Machine Learning services
23mComputer Vision Toolbox - Connected Computer Vision & AI Quality Inspection at Scale!