You Can't Build a New Machine Learning Product with Old DevOps

If you’re trying to build a product that contains an ML model, traditional Developer Operations (DevOps) aren’t ideal for organizing your work.

Here’s a table showing why that is,

along with some possible solutions.

Further reading:

• The Full Stack Deep Learning course
• Hidden Technical Debt in Machine Learning Systems
• Rules of Machine Learning: Best Practices for ML Engineering
• What makes ImageNet good for transfer learning?