The Strangler Fig Pattern in Practice: Retiring a Legacy Pricing Engine Without a Big-Bang Rewrite
How a facade-and-redirect strategy let a 12-year-old pricing monolith get replaced piece by piece, with checkout never going dark.

Full-Stack Developer · E-Commerce & PIM Specialist
15+ years in enterprise e-commerce, Java backends, and modern frontend frameworks. Specialized in marketplace platforms (SAP Hybris, Intershop ICM), microservices with Spring Boot and Quarkus, product data pipelines, and cloud-native services. I integrate AI tooling — including Claude — into my daily workflow to sharpen architecture decisions, accelerate delivery, and write better software.
"I believe in clear words and honest work."
In-depth technical write-ups from real production architectures.
How a facade-and-redirect strategy let a 12-year-old pricing monolith get replaced piece by piece, with checkout never going dark.
A PIM export is a multi-gigabyte NDJSON file. The team needs 200 enriched products from an Excel list, in minutes, reliably. Here is how I built the pipeline: SFTP with smart caching, byte-level GTIN filtering with leading-zero normalisation, memoised category tree resolution, and threaded enrichment in a single pass with orjson.
A payment provider required mutual TLS. What looked like 'just add a certificate' turned into a trust-store gotcha that could break every other HTTPS call in the JVM, two competing secret-sync mechanisms quietly coexisting in the same cluster, and a keystore password Azure Key Vault throws away by design.