Database and API performance work on a commerce and CRM platform, alongside backend team management.
My role: Engineering Manager / backend architecture owner
Scope: NestJS modules, Prisma repositories, PostgreSQL, Redis, RabbitMQ, Nginx, delivery process
The problem
Customer-facing CRM and commerce workflows needed faster responses while the team continued to ship ticketing, automation, and chat features.
My contribution
I tightened Prisma read paths, bounded queries, added indexes and caching, and supported production monitoring. Since early 2024, I have also managed backend engineers through delivery planning, reviews, hiring, and mentoring.
Engineering decision
Tighten the read path before changing the architecture
Approach
Use bounded list queries and explicit Prisma fields, then add cache entries with clear ownership. Keep list and detail data needs separate so the UI contract survives the performance change.
Alternative and tradeoff
A broad rewrite would increase the regression surface while the team was shipping. Caching the original broad query alone would leave the expensive miss path and stale-data risk intact.
Reviewing the behavior
Compare the same request shape before and after: filters, authorization, ordering, page size, payload, and latency. Watch the uncached path as well as cache hits; verify the rollback path before release.
Measurement and scope
Over approximately six months, database/API timing tests showed a 40% reduction in average API latency. UptimeRobot, internal CRM incident records, and monitoring tracked 99.9% platform uptime. These are my reported results from private company records, not an independently audited SLA or a measurement of every endpoint.