Paweł D.-Backend Engineer • Payments & Event-Driven Systems

Independent contractor across fintech, commerce and marketing platforms.

Money-critical services on a platform processing millions of approved transactions per month

Payments and payouts — idempotent request handling, exactly-once processing, complete audit trail for every state transition

Ledger and reconciliation — amounts in minor units, automated matching of internal records against provider settlements

KYC and onboarding flows, plus reporting and data pipelines over transactional data

Overlapping hours with the US team; covered integration, ledger, compliance and reporting rather than a narrow slice

Only engineer on the product — a one-person engineering department owning architecture, AWS infrastructure, delivery and production reliability

Cut key page load time from 5s to 2s while traffic grew 10x

Brought database utilisation down from sustained at the instance limit to ~30% across that same 10x growth

Diagnosed and removed lock contention (6s row-level lock) that made horizontal scaling ineffective — ten instances could not clear 2 webhooks/s until the contention was fixed

Fixed event publication happening before transaction commit, eliminating cross-service race conditions on stale reads

Rewrote a core sales component from PHP to Go and shipped it to production — chosen for the concurrency model after the lock contention work above, where PHP's shared-nothing worker model was the wrong fit

Redesigned order processing as a 7-step asynchronous chain over a slow, unreliable third-party event source — idempotency at every hop, retries, duplicate and out-of-order handling

Worked directly with founders — turned business goals into scope, estimates and trade-offs, explained in non-technical terms

Owned the full SDLC: CI/CD, monitoring and alerting (CloudWatch), incident response, post-incident hardening

Beyond assigned scope: proposed and shipped the platform's first production LLM features

Laravel and Symfony backend for a platform serving 2.5M+ monthly active users

WMS integration processing 150,000+ daily orders

Redis caching and distributed locks eliminating race conditions during flash sales with 50,000+ concurrent users

AWS — ECS, RDS, Lambda, CloudWatch and Auto Scaling absorbing 3x promotional traffic spikes

Newsletter platform for 500,000+ subscribers, delivering 2–3M emails monthly with real-time tracking

Asynchronous PHP with Temporal and Swoole — campaign dispatch time cut by 60%

Production work across MySQL/MariaDB, PostgreSQL and MSSQL — query and index strategy, transaction isolation levels, deadlock diagnosis, large-dataset processing with a minimal memory footprint

Behat (E2E) and PHPUnit, 60%+ coverage maintained, TDD for new features

Shipped four production LLM features — document data extraction, support automation, content generation

Built the pipeline around the model, not just the API call — queued async processing (RabbitMQ / Temporal), structured output with schema validation, retries and fallbacks so model failures never break the business flow

Regression test suites for prompt changes; prompts versioned as production code. Cost and latency controlled through response caching, per-task model selection and token budgeting

Daily AI-assisted development workflow (Claude Code) — agentic refactoring on legacy codebases, test generation, automated first-pass review