Started deep in .NET / Blazor enterprise systems. Learned that code is easy, maintainability is hard so I focus on clean architecture, scalable batch systems and docs that survive handoffs.
Now applying the same discipline to Rust services + React frontends + cloud infra, with AI as engineering leverage.
- Currently: batch processing + dynamic expression libraries.
- Exploring: docs-as-code (Markdown + Mermaid).
- Open-source: I like to contribute to open-source projects, usually to tools that I use, such as Firefly Pico, PowerToys Run plugins, WFHelper but if there are projects to contribute to, why not?
- Beyond code: family + homelab tinkering.
| Design | Build | Operate |
|---|---|---|
| Rules → domain models, explicit contracts | .NET / Rust backends, React frontends | Observable jobs, refactors, runbooks |
| You have… | I bring… |
|---|---|
| Spaghetti rules nobody touches | Modeling + incremental cleanup |
| Batch jobs failing at 3am | Idempotent pipelines + observability |
| AI experiments stuck in demo | Harness + skills + verification that ships |
- Harnesses: isolated, reviewable work units with plan + todo. Auditable by design.
- Skills: brainstorm → plan → TDD → debug systematically → verify before merge.
- Plugins & protocols: code, docs and tools exposed as structured context. No copy-paste.
- Memory: decisions and gotchas stay in the repo (ADRs, lessons), not in chat.
- Humans decide: AI proposes, tests argue, humans merge.
I use AI to raise the quality bar, not to skip it.
- Problem & constraints first: understand the business rule, its edge cases and what "done" means before any code.
- Domain modeling: entities, value objects and aggregates (DDD building blocks) business language in code, not CRUD shapes.
- Architecture: ports & adapters with clean layers domain at the center, infrastructure at the edges, dependencies pointing inward.
- Patterns where they pay: Strategy for interchangeable business rules, Decorator for cross-cutting concerns, CQRS when reads and writes scale differently, idempotent steps for batch jobs.
- Decisions on record: one ADR per meaningful choice, TDD as executable spec, docs-as-code so the next dev doesn't need me on call.
Simple beats clever. If the next maintainer can't follow it, it's not finished.
Where I want to dig deeper:
- From PowerToys Run plugins → upstream microsoft/PowerToys contributions.
- Blazor component ecosystem closest to my enterprise background.
- Firefly III ecosystem finance tooling I actually use.
- Rust CLIs around batch/finance workflows where my current learning compounds.




