Piotr MuchaPrincipal Software Engineer FairyDeck EN / PL Discuss your system
Menu

Featured case · Anonymous automotive ERP

Production modernization

Modernizing a critical ERP without stopping sales, purchasing or warehouse operations.

Production OMS and SCM now own core workflows, legacy delivery is reversible and old and new systems share one observability layer.

20k
orders processed each month
3M
SKU in the ERP environment
40 sec
maximum time for a reversible legacy deployment
No downtime
production database migration with no downtime

01

Modernize the operating model, not just the codebase.

The legacy environment used separate databases, direct cross-system writes and files instead of explicit contracts. Deployments depended on one person, failures reached users before engineers and large supplier orders required hours of manual work.

The modernization therefore began with stability, observability and reversible delivery. Complete processes then moved into domains with explicit ownership instead of being split indefinitely between old and new systems.

02

A critical system that could not be paused or safely rewritten.

Initial state

Undocumented responsibilities and conflicting data

Several legacy applications kept their own product data and communicated through database access and files. Operational decisions were difficult to audit and production failures were often first reported by end users.

Constraints

Every step had to preserve business continuity

  • Sales, purchasing, warehouse work and financial documents had to remain available.
  • Domain knowledge was incomplete and distributed across people and code.
  • Tight coupling made endpoint-by-endpoint extraction less safe than moving complete processes.
  • Every deployment and data transition needed an observable rollback path.

Why not a rewrite

Unknown business rules make a big-bang replacement unsafe

A complete rewrite would combine the highest technical uncertainty with the largest possible release. Process-level migration let the team learn from production while keeping one clear owner for each responsibility.

Personal ownership

Architecture remained connected to implementation and operations.

I combined system architecture, critical hands-on work and production ownership throughout the transition.

Modernization and domain architecture
I set the migration strategy, domain boundaries, data ownership and communication model for the new ERP.
OMS, SCM and supporting services
I designed and implemented OMS end to end. I designed SCM, built its frontend and delivered later changes; another developer implemented part of the initial backend.
Safe production change
I owned AWS and IaC decisions, legacy deployment automation, the observability platform and the production database migration.

PIM, WMS, Pricing Engine, Marketplace Integration Service and central BI are still being implemented or designed. They are described as current work, not as completed outcomes.

Key decisions

Reduce operational risk before increasing the rate of change.

The architecture followed the business process and selected the smallest reversible step that transferred clear responsibility.

  1. 01

    Move complete processes

    OMS and SCM took over coherent workflows instead of leaving hidden ownership split across implementations.

  2. 02

    Stabilize before extraction

    Automated deployments, rollback, monitoring and backups made the transition safer to operate.

  3. 03

    Match communication to the operation

    REST supports direct responses; Kafka and events propagate asynchronous changes between domains.

  4. 04

    Migrate data under replication

    Dump and restore established the base state, then replication was observed before the controlled RDS cutover.

Delivery model

A sequence of operationally complete changes.

  1. 01

    Stabilize and observe

    Secure delivery, access, recovery and production signals around the systems the business already depends on.

  2. 02

    Make ownership explicit

    Map critical processes, truth sources and hidden rules before choosing a new domain boundary.

  3. 03

    Extract the complete workflow

    Build the new component, integrate it through explicit contracts and switch a complete business responsibility.

  4. 04

    Verify and retire

    Measure the production result and remove the legacy responsibility only after every dependency has moved.

Verified production outcomes

Less manual work, shorter risk windows and earlier signals.

Several hours → several minutes
for an order containing about 2,000 parts
About one hour → max. 40 seconds
maximum reversible deployment time, down from about one hour
User report → alert within seconds
across eight legacy systems, six new ERP parts and their infrastructure
12 GB and 59 GB databases · no downtime
for 12 GB and 59 GB production databases moved to AWS RDS

The order-cycle improvement belongs to the combined OMS and SCM flow. Supplier selection is the main SCM optimization.

Current state

Production responsibilities and work still in progress stay separate.

The modernization is delivering value before the old ERP is fully retired. Remaining domains are presented through their problem and architecture, not future results.

What this proves

Principal-level ownership inside a live, critical environment.

I can enter a weakly documented system, stabilize what the business cannot stop and translate operational constraints into domain boundaries, critical implementation and measurable production change.

Automating catalog updates for nearly 10 million automotive products.

Supplier data was incomplete and inconsistent, while each catalog update required days of developer work. I designed an automated architecture that processes weekly TecAlliance deltas after publication and keeps a standardized catalog of nearly 10 million products current.

Modernizing a marketing automation platform processing up to about 45,000 events per minute.

At roughly 1,000 integrated stores, disk-based event processing and server-by-server scaling were reaching reliability and cost limits. The platform moved to queues and scalable workers while critical PHP and Java components were modernized incrementally.

Your modernization

Is your ERP too important to stop and too risky to leave unchanged?

We can start with the critical process, the unacceptable failure and the responsibility that must move first.

Discuss a critical-system modernization

Back to selected work