Backend systems that don't break under load.

systems · automation · web · strategy

I help startups and small teams build, fix, and scale APIs, infrastructure, deployments, and online presence.

Get a system audit

Common problems I solve

  • APIs that are slow or unstable
  • Deployments that break in production
  • Messy backend architectures
  • Manual processes that should be automated
  • Systems that don't scale

How I work

Fixed-scope engagements with clear deliverables — not open-ended consulting.

Backend Fix Sprint

1–2 weeks
  • System audit
  • Performance fixes
  • Stability improvements
  • Deploy and verify changes

API / Infrastructure Setup

project-based
  • API design and implementation
  • CI/CD pipelines
  • Docker and cloud setup

Scaling & Optimization

project-based
  • Bottleneck analysis
  • Performance tuning
  • Architecture cleanup
Fix my backend →

Selected work

Freelance and early-stage projects. Results from production systems.

API latency reduction in production system p99: ~900ms → ~180ms Production SaaS API — query optimization, caching, safer deployments.

Context

A backend API powering a small production SaaS was experiencing inconsistent performance under load. p95/p99 response times were degrading during peak traffic, impacting user experience and increasing server costs.

REST architecture, relational database, multiple external service calls.

Problem

  • p99 latency reached ~900ms under moderate load
  • Database queries not optimized for real-world usage
  • N+1 query patterns in core endpoints
  • No caching for frequently requested resources
  • Risky deployments with no clear rollback strategy

Approach

End-to-end analysis focused on runtime behavior, not just code structure.

  • Profiling API endpoints under simulated load
  • Query optimization and removal of N+1 patterns
  • Selective caching for high-frequency reads
  • Reduced external service calls in request path
  • Improved deployment flow with safe rollback

Results

  • p99 latency reduced from ~900ms to ~180ms
  • Significant reduction in database load at peak
  • More stable performance during traffic spikes
  • Safer, faster deployments with rollback support

Stack

Go / Node.js, PostgreSQL, Docker, AWS, CI/CD

Most performance issues were not caused by infrastructure limits, but by request path inefficiencies and database access patterns.

What I'd look for next time

  • Reduce external dependencies in hot paths
  • Better observability — metrics per endpoint
  • Pre-emptive scaling strategy for traffic spikes
Systems & tooling experiments on GitHub

Stack

Systems
Go, Python, Node.js
Infra
Docker, Kubernetes, Linux
Cloud
AWS, CI/CD, Terraform

Let's review your infrastructure.

Get a system audit
hello@szpila.tech
  • Usually respond within 24h
  • 1–2 clients at a time
  • Remote · EU timezone