Order Management Systems at dwpbank

The context

dwpbank (Deutsche WertpapierService Bank AG) provides securities services for hundreds of German banks. The Order Management System (OMS) processes retail and institutional securities orders at scale.

My role

Dev Lead, Fullstack dev on core OMS modules: Spring Boot services, Kafka event streams, JPA persistence, Angular frontend, GitLab CI/CD, AWS deployment.

What I shipped

Day-to-day work spanned event-driven workflows on the backend, frontend modules in Angular, and integration improvements across the connectivity layer — all delivered through the team's CI/CD pipeline against the observability, audit, and uptime expectations of a regulated financial environment.

What this means for AI work

The discipline of shipping into a regulated, audited, high-availability codebase translates directly to integrating AI features into production systems where uptime and observability are non-negotiable. The same instincts — backward-compatible changes, idempotent handlers, careful rollouts, real alerting — are what separate AI features that survive in production from prototypes that don't.

Related work

cravingtoolkit.com — RAG and AI content pipeline — production AI in a YMYL domain.
Multi-agent automation infrastructure — self-hosted LLM agents with provider routing and cost discipline.

Full architecture write-up coming soon.

Happy to walk through this in detail on a call — hello@jakubhavelka.dev.