Key Results

  • At least 10 hours a month saved in Engineering Time
  • Proactively reach out to customers within 1 or 2 hours of discovering an error
  • Accelerated error identification and resolution
  • Seamless integration across various platforms
  • Efficient workflow through customizable features
  • Immediate alerting and quick triaging


Python (Django), Golang


Error Monitoring

By filling out this form, you agree to our privacy policy. This form is protected by reCAPTCHA and Google's Privacy Policy and Terms of Service apply.


Share on Twitter.
Share on Facebook.
Share on LinkedIn.

How Baseten Saves 10 Hours a Month through Accelerated Error Identification and Resolution

About Baseten

Headquartered in San Francisco, Baseten is a platform that enables its users to develop, deploy, and test ML models in production fast and reliably. For engineers and ML practitioners who would otherwise be mired in layers of Kubernetes and cloud configuration, Baseten allows them to focus on the things they want to do — actually work on their models and develop their applications.

As the foundation behind large, mission-critical applications, Baseten requires constant visibility across their platform. Missing out on notifications for errors or system issues could compromise not only their operations but also user trust in their platform.

In-Depth Error Context Leads to Quick Issue Resolution

In an environment where every hour spent on detecting issues could mean a loss in users, Baseten required actionable error details. And, as a hosting provider offering several services, they needed a way to detect the many edge cases silently affecting their platform.

As a fast-moving startup that deploys at least once a day, it’s crucial that we have the tools to be responsive to customers and have insight when there are bugs or regressions on the platform.

Sid Shanker, Backend Engineer

After hearing about Sentry’s robust error monitoring capabilities, Baseten decided to test it out. Immediately, they saw an increase in the error details they could access. Previously, the Baseten team required days to track down any non-reported issues. Now, by leveraging stack traces, they were able to detect and retrieve the source code of these issues within several hours. And, if any platform updates downgraded the customer experience, Baseten was able to identify the errors, rollback changes, and resolve the issues without spending an extra day of turnaround time to add the requisite logging.


Multi-Project Versatility Supports Hypergrowth

With a rapidly expanding team working on refining and releasing new product features as soon as possible, Baseten needed an error monitoring solution that could easily integrate with both their existing code base and any new projects.

Baseten is a hosting platform that has a few moving pieces, so it’s important that we have tools that are easy to integrate into different services in a polyglot environment.

Sid Shanker, Backend Engineer

While launching a new service to handle authentication for model inference, they relied on Sentry to create an error-free experience. Unlike their main Python-based web application, this new service was written in Golang. Due to Sentry’s broad SDK support and detailed documentation, the Baseten team gained confidence that their new service was handling model inferences without errors.

Sentry is a key part of operating our new service — the additional context we pass to Sentry on each call has helped us solve pretty tricky bugs, saving countless engineering hours.

Sid Shanker, Backend Engineer

Customizable Features Ensure Smooth Workflow

As a hosting platform for technical ML models, even minor glitches can escalate into major setbacks. This is especially the case when it’s unclear whether the platform or the models themselves are experiencing errors. By customizing Sentry for their needs, Baseten can focus on detecting the issues that they can resolve and maintain user confidence.

Baseten has made use of Sentry’s broad integration options to create a real-time alert system as part of their internal communications on Slack. This setup serves as a hub for immediate updates, fulfilling an integral role in their forward-deployed engineering workflows. With alerts directly channeled into Slack, on-call engineers can swiftly triage novel issues by correlating customer-reported problems with the alerts channel.

Baseten also leverages another significant Sentry customization: adding extra data to events. By providing the context needed to understand their bugs, the engineers can save hours of work on manually logging specific issues.

Future-Proofing: What’s Next for Baseten

With technological advances in machine learning occurring at breakneck speed, Baseten understands the need to stay ahead of the curve. As the company expands to new services and product areas, Sentry’s scalability ensures that Baseten can maintain the same level of oversight and responsiveness across all of their platforms.

The relationship between Baseten and Sentry is a testament to how the right tools can amplify a company’s operational efficiency. Whether it’s through detailed error context, multi-platform support, or customizable features, Baseten has optimized its platform for ensuring user trust. By continuing to adopt and adapt Sentry to its future use cases, Baseten can continue innovating while providing a user-centric platform.

A better experience for your users. An easier life for your developers.

© 2024 • Sentry is a registered Trademark
of Functional Software, Inc.