Platform solves model building. We help with the rest.
At this stage we look into project requirements set forth by our clients. We also establish what data collections should look like for competitions and set up other details related to project scope.
The requirements may include skilling the workforce on the details of platform, employing a competition, cloud deployment, and data privacy specifics.
There are two ways our clients can go about this stage. Either the models are developed by client's using platform internally or we create competitions.
Either an internal workforce is trained to use platform, or we arrange a competition that our fellows participate in. For any project we require two collections of data - a hidden set for testing and validation and a training set for use on platform. Once the platform model has reached a sufficient level of accuracy, it is ready to be deployed.
Once the model is ready, we can also help our clients set up CI/CD environments for the deployment of model-based microservices.
Platform models are ready for deployment via the fastai toolchain. While designing environments, We adopt ML-Ops best practices at this stage and ensure the model APIs are encapsulated effectively to make deployments and service calls easier.