Batch jobs ran overnight, so by the time anyone saw a failure the window had closed and the day already started behind. I moved the critical paths to streaming, made retries safe to repeat, and turned a nightly gamble into something that just quietly kept up.

Right-sized clusters and added autoscaling; the monthly bill dropped 50% 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.

Replaced manual releases with a gated pipeline; deploy time dropped 4x and rollbacks became routine.

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.

Replayed the write-ahead log and reconciled balances; zero customer-visible discrepancies.

Two teams were blocked on the same shared bottleneck and each was waiting for the other to move first. I designed the contract between them and the migration path to reach it, so both could ship on their own schedule instead of renegotiating scope at every release.

Decomposed the monolith behind a strangler proxy into 9 services — no maintenance window needed.