I build custom web applications and business systems.

Since joining full-time in 2019, I have worked across two generations of lending software: a modern multi-tenant platform and the older application it is gradually replacing. I started mainly in frontend development, became one of the main contributors to the component library that grew into the modern platform's design system, and designed interfaces for information-heavy loan workflows. I later moved into backend architecture and designed the original extension model used to customize loan processing for individual banks.

I remain heavily involved in the modern platform's development. My recent work includes core loan-processing workflows that coordinate automated background checks and manual review actions that can proceed in parallel, with safe pause, resume, and retry behavior. I also help other developers shape larger features through design discussions, mentoring, and detailed reviews.

At the same time, I have technical ownership of the older lending application after its previous team left. I learned the large PHP codebase from its code, data, limited documentation, and live behavior while continuing to deliver features and fixes requested by the bank. I investigate live issues, coordinate on-premises releases and maintenance windows, and work directly with company leadership and the banking client around priorities and rollouts.

My largest recent responsibility has been preparing both applications for a banking client's move to a new core-banking system. I placed the new integration work in the modern platform while keeping the older application's existing workflows operational, then took over a customer-data migration involving millions of records. Testing changes is slow: restoring the legacy database can take most of a day, and a full migration run takes several hours. Because a problem discovered late could consume an entire testing cycle, I replaced a direct update script with a staged process that can plan, validate, report, execute, stop, and resume safely. Three full trial migrations against restored bank data each completed in roughly three hours, with stop-and-resume behavior verified.

After its previous developer became unavailable, I took over a retail platform used by more than 20 Norwegian shops. With no technical handover, I had to reconstruct earlier decisions while fixing live issues and continuing feature development.

When an accidental user action removed the recorded locations of stored tire sets, I wrote a dedicated recovery script using the request and audit logs I had introduced, together with timestamps, vehicle registrations, movement history, and tire attributes. It reconstructed nearly every location, avoiding a large manual recovery effort and the risk of shop staff being unable to find customers' stored tires when they arrived.

For Norway's SAF-T cash-register requirements, I led the implementation from the official specification through validation by the authorities and an external specialist. I later moved the fiscal and reporting features into a separate multi-tenant application. For an accounting integration, I worked directly with the client's CEO and accountant to map retail activity into balanced ledger entries, replacing manual entry of every Z report with automatically generated vouchers.

To prepare the system for failures, I compared three high-availability designs across recovery time, possible data loss, failure scenarios, cost, and maintenance. I recommended the managed option because its reliability and lower maintenance justified the higher infrastructure cost, then implemented the selected design with Terraform. During an agreed smoke test while the shops were open, I stopped an application server and forced a database failover without shop users noticing an interruption.

The email platform I designed and built sends through its own pool of mail servers rather than forwarding messages through an external delivery provider. Its scheduler assigns messages across servers and IP addresses while respecting limits for senders, domains, and destination email providers.

Messages have clear states and can be retried without losing the previous SMTP result or consuming the same sending capacity twice. The platform also manages automated gradual sender warm-up, domain setup, bounce and unsubscribe handling, engagement tracking, administration, and monitoring. I also set up monitoring and centralized logging with Grafana, Prometheus, and Loki, covering mail throughput, queue depth, processing delays, message states, and delivery outcomes.