Workflow
TIM Studio is well equipped so that you can explore your data, iterate over experiments, and then turn easily to regular (production) forecasting.
The flow
Users in TIM Studio can rely on automation and user-friendly functionality for both phases of MLDevOps.
Typical flow of a User, from uploading data to a production, looks as follows.
| Step | Activity | MLDevOps phase |
|---|---|---|
| 1 | upload Dataset, explore it | |
| 2 | back-test / iterate over experiments, get insights into models built by TIM | ML |
| 3 | select Iteration of Experiment with the best results and data to be used in production and activate it as Production setup | Ops (Release) |
| 4 | use your Production setup in regular (production) forecasting | Ops |
Term production refers to use of already established settings and data without the need to further experiment, inspection of model details etc. This kind of use can be done manually within TIM Studio interface, or automated thanks to TIM Studio API when implemented in your application stack.
Artifacts
TIM Studio organizes its artifacts in hierarchical structure as follows:
- Datasets
- Use Cases
- Experiments
- Iterations (of Experiment)
- Production setup (derived from one of Iterations and Dataset)
- Experiments