In this article, Dan Mindru, a frontend developer and designer, uses Sentry's Trace View to identify performance bottlenecks in his API call. He discovers that the endpoint is slow due to multiple HTTP calls, file I/O, third-party calls, and database queries. To optimize the performance, he sets up custom instrumentation using Sentry's methods to track operations, such as file I/O and HTTPS calls. He then identifies long-running spans, waterfall spans that wait for each other, and inefficient span order that have to wait on each other. By parallelizing decoupled spans, optimizing image processing, changing the Gen AI model, moving the Gen AI span to start, and reordering independent spans, he is able to shave off 22.3 seconds of load time from his API call. The article concludes with a discussion of future optimization opportunities and emphasizes the importance of setting up tracking to identify performance bottlenecks.