Projects
A public-safe map of the work I keep returning to. Each area below says what you can actually read on this site and links to the published writing behind it. Private implementation detail, internal systems, and anything touching patient care are left out on purpose.
AI-native operations
Designing workflows where agents, specs, tests, and humans share a clean operating surface. The idea is simple: an agent should be able to pick up a task, read the rules, make a bounded change, and leave a record that a human can check without re-reading the whole history.
What you can read here: how a project starts, how agent roles are split so the work stays legible, and how one small change moves from planning through tests to delivery.
- Day Zero: how my AI projects actually begin, a field note from auditing 94 project directories and turning project starts into a repeatable ritual.
- How I run AI agents, on using one model for plan mode, another for bounded execution, and a slim harness to keep roles narrow.
- The layers behind one website revision, a case study of one change moving through planning, orchestration, specialist agents, tests, and review.
Clinical systems
Building safer, simpler systems for prevention-oriented kidney care without exposing patient details. Medicine taught me that a good system makes the right action easier and a bad one asks people to compensate forever; that lens shapes everything else on this page.
What you can read here: the systems perspective, not the clinical tooling. This site does not publish patient information, practice internals, or clinical software details. My clinical work is with The Kidney Experts, and questions about care belong with the practice, not this personal site.
Cloud infrastructure
Practical deployment, observability, and automation patterns for small teams that need reliable systems. The goal is infrastructure a small team can actually understand and operate, using modern cloud primitives without the complexity bloat.
What you can read here: a written guide and an interactive companion on keeping secrets out of source code, which is where a lot of small-team cloud trouble starts.
- Google Secret Manager: stop hardcoding API keys, a migration and rotation path from keys in code to versioned, IAM-controlled storage.
- Google Secret Manager, explained visually, a hands-on guide covering secrets and versions, what
latestpoints at, and the create, rotate, disable, and destroy lifecycle.
How to read this page
This is a map of themes, not a status board. It does not say what is underway, what is finished, or what comes next, and it should not be read as a description of any internal system. For the published notes themselves, go to Writing. For the themes I keep returning to, see Now. For public, professional contact, use the Contact page.