Selected work
Architecture decisions proven in production.
These are not technology showcases. Each case starts with a business process that had to keep running and makes my responsibility, the constraints and the verified result explicit.
Client names are used only where publication is approved. The automotive ERP remains anonymous.
Piotr MuchaPrincipal Software Engineer · FairyDeck
Featured case · Anonymous automotive ERP
Modernizing a critical ERP without stopping sales, purchasing or warehouse operations.
A process-by-process modernization of a tightly coupled ERP, with production OMS and SCM, reversible delivery and one observability layer across legacy and new systems.
Personal ownership: I own the modernization architecture, critical implementation paths and production safety. Work implemented by other developers is identified explicitly.
Read the ERP modernization case- 20k
- orders processed each month
- 3M
- SKU in the ERP environment
- Several hours → several minutes
- for an order containing about 2,000 parts
- About one hour → max. 40 seconds
- maximum time for a reversible legacy deployment
Four additional production stories
Different systems. The same standard of evidence.
-
02
Published production experience
Modernizing a marketing automation platform processing up to about 45,000 events per minute.
A marketing automation platform moved from disk-based event handling to queues and scalable workers while critical PHP and Java applications stayed online.
Up to about 45,000 events per minute, with no event loss during Black Friday peaks.Hands-on event-processing migration, legacy modernization and e-commerce integrations.
Read the edrone case -
03
Published production experience
Normalizing European PSD2 bank integrations for an Account-to-Account platform handling hundreds of thousands of transactions each month.
European bank-specific APIs, statuses and failures were normalized behind one Account-to-Account payment process.
Production integrations supporting a platform handling hundreds of thousands of transactions each month.Design and implementation of individual bank integrations, including status and error normalization.
Read the Volt.io case -
04
Published production experience
Removing customer lookup from support calls on an e-commerce platform serving more than 20,000 stores.
Telephony, customer administration and support history were connected so the right context was ready before an agent answered a call.
A lookup step that could consume up to 60% of handling time was practically eliminated.Hands-on delivery of the support-context workflow and a separate PrestaShop migration application.
Read the Shoper case -
05
Architecture delivered to production
Automating catalog updates for nearly 10 million automotive products.
A developer-dependent catalog update process was replaced with automated processing of standardized automotive product data.
Weekly deltas for a catalog of nearly 10 million products are processed after they become available.I designed the architecture and data flow; other developers implemented the solution.
Read the TecAlliance case
What the evidence covers
Scale is useful only when its context stays attached.
Business continuity during ERP change
Sales, purchasing and warehouse operations remained online while complete processes moved out of legacy.
Reliable event processing at peak load
Queues and scalable workers replaced file-based handling without a big-bang rewrite.
Regulated integrations with explicit failure semantics
Bank-specific contracts became one internal process with Mutual TLS, certificates and controlled retry behavior.
Operational context prepared before the call
A support workflow was redesigned around the agent's real decision path, not another isolated tool.
Architecture that removes recurring manual work
A large catalog pipeline became automatic while implementation ownership remained transparent.
Your system
Which critical process is currently too risky to change?
Bring the business process, the operational constraint and the failure you cannot accept. We will use the first conversation to determine a safe next step.
Discuss your system