With just two lines of code, you can train, test, serialize, and containerize your application. Qwak CI enables you to get your hands on fully-loaded production-grade models with zero effort.
All you need is one click (or API command) to manage live end-points from the containerized version of your model.
Analytics is part of Qwak’s off-the-shelf product offering. No additional effort is required to track your model's data; all you have to do is decide(during the CI process) if you would like to track the model data. We will do the rest ;)
Qwak also enables you to track model metrics, such as CPU, GPU, memory, throughput, etc. These metrics are integral to the model’s health and should be used in conjunction with the model’s data monitoring.
Traditionally, data scientists write the Features logic for training operations, which need to be “reverse” engineered by the engineers who consume the models once they are live, for the inference phase.
The feature store eliminates this stage by allowing the DS to define features in a single way for “offline” and “online” operations. Doing so empowers data scientists and engineers to better collaborate on features so that data scientists can add and remove features, while remaining in sync with the engineering teams.