The Stack Overflow Podcast Improving error monitoring with AI
29 snips
Mar 18, 2025 Tillman Elser, Engineering Manager for AI and ML at Sentry, shares his journey from economic research to cutting-edge tech. He dives into how semantic embeddings, especially using BERT, revolutionize error monitoring by transforming complex stack traces into actionable insights. The discussion also sheds light on enhancing customer engagement in data labeling akin to an escape room challenge. Finally, Tillman highlights breakthroughs in error monitoring tools, showcasing the impact of generative AI in ensuring security and privacy for developers.
AI Snips
Chapters
Transcript
Episode notes
From Economics to Tech
- Tillman Elser's career path began in economics at the Federal Reserve Board.
- He transitioned to tech, focusing on price forecasting and ad tech before joining Sentry.
AI-Powered Error Grouping
- Sentry uses embeddings models to group similar stack traces, improving error aggregation.
- This helps developers quickly identify and address recurring issues in their code.
Filtering Stack Trace Noise
- Stack traces combine application code with lower-level system details.
- Sentry filters out system-level noise to focus on application logic for better error grouping.