ObservabilityModel Monitoring

Collect store, and analyze your model's data centrally to fully monitor inference, feedback and baseline data.

Model Monitoring Overview

Qwak Model Monitoring is a fully managed data lake that comes as an off-the-shelf product with the Qwak Platform. With Qwak Model Monitoring you find all your inference, feedback, and baseline data for each model and its different versions.
‍
Model Monitoring data can be accessed directly via the Qwak SQL interface, or via any other query or data processor engine.

The Main Pillars

Data Observability

Track your models production data and compare it to that model’s training and its previous versions.

Audit

Log all your inference data from its inception, at any scale.

Performance Management

Monitor applicative and statistical model performance metrics.

Getting Started

Setting up Qwak Model Monitoring requires only one line of code. Just import the Qwak SDK and wrap the relevant functions (the predict function, in our example) with @analytics. Qwak will automatically track and manage all your data into Model Monitoring.

Start for free

Model Monitoring Use Cases

Track and monitor your models features and predictions

With Model Monitoring every inference is represented by a separate row. You can track the distribution, performance, and changes in your model behavior across times or builds.

Model Feedback

Collect the actual results of the model. Sent results are automatically attached to the relevant predictions data, allowing you to track your model performance metrics in real time.

Retrain

Since you have all the data within the system, you can use the Qwak Model Monitoring data for retraining purposes, which can be fully automated with Qwak Pipeline Orchestrator.

Integrate with other tools

Qwak Model Monitoring data can be integrated with third-party vendors, from Audit tools to data and model monitoring tools, without modifying models code.