- The key to making your ML product a success, it is important that they are data-centric and architecture agnostic. This way, we can quickly experiment with different models on our specific dataset.
The data engineering team owns the data pipeline The ML engineering team owns the FTI (feature, training, and inference) pipelines.
- The training pipelines are scaled vertically by adding more GPUs because training is computationally expensive. Inference pipelines are scaled horizontally, based on the total of number of concurrent user requests.