Upstream code changes, breakdowns, human interference, new clients with different data characteristics or another reason: ML systems can break down for reasons outside your control.
Raymon helps you gain insight in the data your systems process and helps you discover, troubleshoot and remediate failures in production systems.
Text logs are useful, but are insufficient for ML applications. Using our open source library you can log any data type you want to our backend. This opens up a realm of possibilities.
Trace data as it flows through your deployment and inspect intermediate results. Debug issues and develop a feeling of what actual production data currently looks like and what can be improved in your system.
Using our open source data validation library you can easily define what kind of data your ML system expects in flexible and easily constructible data schema’s. When deviations occur, you get notified.
ML systems may serve different clients that may have different data characteristics. Moreover, only some clients may face issues and suffer from degraded performance, while others do not. Using our slice-based monitoring and issue discovery, one client's issues will not get lost in the noise of other, unaffected clients.
Logging data takes just a few lines of code. We offer lightweight, simple Python libraries.
We are not tied to any framework or library. PyTorch, sklearn, numpy, ... No problem!
Run the Raymon backend yourself or use our hosted solution.
OUR PARTNERS IN EXPANSION
Leave your email and we'll keep you posted on what we are building.