Tomasz Kowal
tomaszk
Backend. Postgres apologist. Most of these are things I got tired of explaining.
Bring a schema's table and column names onto one convention, and list the migrations that would get there.
Find writes that should be atomic and aren't — multi-step mutations with no transaction around them.
Given a table's schema and its hottest queries, propose the smallest set of indexes that serves them.
Trace an endpoint's queries and flag loops that issue one query per row, with the join or batch that replaces them.
Check a database migration for locks that block writes, missing backfills, and changes that can't be rolled back.
Read a slow Postgres query and its EXPLAIN ANALYZE output, then say which index would change the plan and why.