The pipeline lost rows quietly under load and we only found out when a customer's numbers didn't add up, which is the worst possible way to find out. I chased it to the root — backpressure, at-least-once handling, and reconciliation that proved nothing fell on the floor rather than hoping it hadn't.

Sharded the pipeline and tuned batching; throughput held at 21k events/sec at steady state.

The database had become the whole company's single point of failure, where one team's slow query was everyone else's outage. I put boundaries around it — read replicas, a caching layer with honest invalidation, and the query budgets that stopped one team from taking down the rest.

Right-sized clusters and added autoscaling; the monthly bill dropped 60% at the same SLOs.

The codebase carried a decade of shortcuts, and every change touched three things nobody had written down, so estimates were fiction. I paid down the debt that actually blocked us — carved clean seams, wrote the missing tests first, and made the scary changes routine.

Parallelized and cached the build; PR feedback that used to take the afternoon now lands in minutes.

A third-party dependency in the middle of a critical path went down and took us with it, because we had quietly trusted it never would. I made the system survive its dependencies — timeouts, fallbacks, and circuit breakers that turned an outage into a degradation nobody outside even noticed.

Replaced manual releases with a gated pipeline; deploy time dropped 2x and rollbacks became routine.