How Duolingo debugs issues 12x faster while reducing alert fatigue and empowering engineers

Duolingo, the world’s leading language-learning application, delivers engaging lessons to millions of learners each month. Their mission is straightforward: create the best education in the world and make it universally accessible.

Behind the scenes, more than 200 microservices and a rapid release cadence—with multiple daily deployments—power these experiences. With such a complex system, Duolingo needed to ensure undetected bugs and slow issue resolution didn’t disrupt the platform or their learners-first mission.

Temporary fixes, fragmented debugging, and productivity bottlenecksTemporary fixes, fragmented debugging, and productivity bottlenecks

Before adopting Sentry, Duolingo relied on a legacy error monitoring tool that wasn’t built to handle their scale. The result was a cumbersome debugging workflow and UI that frustrated engineers and limited productivity.

• Slow debugging: Searching for information could take up to 30 minutes for a single query. Only one engineer could query issues at a time, creating an intense bottleneck for the entire team.
• Noise overload: The tool lacked effective filtering and grouping, flooding engineers with irrelevant data and making it difficult to focus on critical issues.
• Limited search capabilities: Granular error trends, like short-term spikes, were nearly impossible to identify. Engineers often resorted to manual methods or relied on intuition to find root causes.
• Difficult-to-use and siloed expertise: With a difficult-to-use UI, its use was limited to a small group of experienced engineers. This approach created a dependency on “power users,” leaving the rest of the team without the necessary tools or context to contribute effectively.
• Limited context and visibility: The legacy tool provided minimal metadata and poor correlation between errors and their root causes, forcing developers to cobble together context from multiple sources. This constant context switching extended time-to-resolution and reduced productivity.

To top it all off, legacy tool couldn’t fully solve all their issues, leaving critical gaps in their workflow. While the team was quick to deliver temporary fixes, these often served as band-aids; root causes were left unresolved, and the same issues would often pop up again down the road.

Even simple debugging felt like a marathon. Only power users could use the tool to debug issues because others couldn’t understand it easily. We didn’t have the tools to make debugging efficient for everyone,” said David Amin, Staff Site Reliability Engineer at Duolingo. “It wasn’t just about wasting time—it was about not having the tools to give every engineer the confidence and ability to solve issues quickly.

With the growing complexity of their systems and teams, Duolingo demanded a more scalable and developer-friendly approach to error monitoring and debugging.

Why Duolingo Chose Sentry: ease of implementation and faster time to insightWhy Duolingo Chose Sentry: ease of implementation and faster time to insight

After evaluating their options, Duolingo chose Sentry because it addressed their biggest challenges with minimal setup. Its straightforward implementation and developer-focused design made it easy for the team to adopt and start seeing results immediately.

• Migrated 200+ microservices in <2 weeks: Sentry’s APIs, Terraform support, and straightforward configuration allowed them to migrate 200+ microservices in just two weeks.
• Accelerated time to resolution by 12x: With real-time queries, intuitive dashboards, and deep context, engineers could quickly filter noise, visualize trends, and pinpoint issues—all within one tool. Previously, finding a root cause was nearly impossible, often taking hours or even days. Now it’s a matter of minutes.
• Noise Reduction: Sentry’s built-in grouping, filtering, Spike Protection, and ownership rules reduce alert fatigue by cutting through the noise, flagging only critical issues and routing them to the right teams. This helps teams stay focused on the errors that matter most. With this clarity, they’ve also integrated directly with PagerDuty to ensure urgent alerts are pushed for immediate action.
• Advanced Context: With out-of-the-box metadata (like browser, release, and user) and custom tags (like experiment IDs and feature flags), Sentry provided Duolingo with a granular view to debug issues by deployment context or individual users.
• Accessible debugging: Sentry’s intuitive UI has made debugging accessible to all developers, enabling them to take ownership of their code and resolve issues proactively. With its straightforward debugging workflows and integrations with the tools they already use, Sentry has become an essential part of Duolingo’s engineering toolkit.

Since rollout, over 50% of Duolingo’s web and backend engineers–– roughly 240 out of 400— use Sentry nearly every day to manage their releases.

Before, with our previous tool, debugging was limited to a few power users. Sentry has democratized the process—now any engineer can jump in, find insights, and fix issues quickly, without extensive training.

David Amin, Observability Team Lead

Results: faster debugging, improved uptime, and empowered teamsResults: faster debugging, improved uptime, and empowered teams

Duolingo developers now resolve issues in minutes instead of hours. A few key examples:

• Fixing database outagesFixing database outages

  Within weeks of adopting Sentry, Duolingo diagnosed a critical database issue that had previously been difficult to pinpoint. Sentry’s dashboards and rich querying capabilities— complete with custom tags and visibility down to the hour — let their team quickly identify an anomaly in one of their most critical databases. They engaged their cloud provider immediately, reducing incident investigation time significantly.

  “With Sentry getting us to the root cause, we could confidently verify the issue in our database allowing us to immediately page the right people with confidence. That level of clarity was not possible before,” said Maggie Hewitt, Engineering Manager

  Even more notably: for subsequent incidents, Sentry allowed them to confidently rule out the database as the source of errors, helping them direct resources efficiently.

• Spotting and stopping spammers in <10 minutesSpotting and stopping spammers in <10 minutes

  After getting tagged in Slack for a spike in errors, an engineer initially suspected an issue with how start_time was handled.

  

  But after checking the Sentry JSON, they confirmed the code correctly accounted for None values. Reviewing related events, they noticed multiple session end events for the same user ID—each with a different session ID and no answered questions. All signs pointed to bot activity.

  Within nine minutes, Sentry provided the context needed to quickly debug the issue, confirm the botting, and take action before it could impact the system.

  

Sentry is one of the few tools that directly links errors to the source code. I’ve seen engineers referencing specific Sentry errors pointing to the exact line of code—we’re all fixing issues faster, and the overall developer experience has improved significantly.

Tim Kirdy, Senior Backend Engineer

• Proactive alertingProactive alerting

  Sentry’s anomaly detection added a new layer of observability to Duolingo’s stack. During an incident where traditional alerts had failed to fire, Sentry flagged a subtle issue in their monolith deploy that might have gone unnoticed. This allowed the monolith team to act quickly, resolving the issue before it escalated.

  Sentry has empowered Duolingo’s developers to take proactive ownership of their code, transforming how teams handle debugging and issue resolution. By democratizing Sentry across the organization, the Observability team enabled individual developers to identify and resolve issues independently—often before the Observability team was even aware of them. This shift eliminated the reliance on centralized command and fostered a “You build it, you run it” culture. Engineers now use Sentry directly to debug and triage their services, enabling faster response times, better accountability, and more efficient workflows.

  Sentry has saved years of my life by making it faster to find trends and errors. What used to take hours of going between tools now happens instantly.

  Maggie Hewitt, Engineering Manager

What’s Next?What’s Next?

With a department-wide focus on improving the user experience and client-side observability, Duolingo is expanding its use of Sentry to their iOS and Android apps, where debugging has historically been slower and more fragmented. By integrating Sentry across both frontend and backend systems, the team hopes to eliminate silos and improve fullstack visibility.

“We’ve seen what Sentry can do for our backend systems,” said Tim Kirdy, Senior Backend Engineer “Now we’re excited to bring that same clarity and speed to our mobile apps.”

Key ResultsKey Results

• Migration Success: 200+ microservices moved in 2 weeks, with minimal disruption.
• Adoption: Over 50% of engineers actively use Sentry daily.
• Impact: Debugging time reduced by 2-3x in key workflows, with fewer silos and better team participation.
• Maintain 99.9% uptime: By enabling faster issue detection and providing the context needed to address root causes, Sentry helps Duolingo fix errors before they escalate into downtime. This has been crucial in maintaining Duolingo’s stringent 99.9% uptime goals.