Specialist engagement
Engineering workflow systems.
Reduce delivery friction without lowering the bar. I design the environments, operating conventions and guardrails that let engineers move quickly and leave a trail another person can understand.
The work
Tools are implementation detail. Repeatability is the product.
For teams, this means safer AI adoption, faster incident handover and environments that new engineers can reproduce. For individuals, it means a workspace built around the work you actually do rather than somebody else's dotfiles.
- Team engagements
- Workflow audit, reference environment, guardrails, documentation and adoption support.
- Individual engagements
- Focused setup, workflow design and a maintainable configuration you own.
- Handover
- Source-controlled configuration, decision record, runbook and a verified rebuild path.
Demonstrations
One real delivery loop. Three focused supporting demonstrations.
The lead recording is an authentic workflow performed against a disposable synthetic repository. The shorter supporting clips use simulated terminal output for repeatability. All media is silent, loads only when requested and includes a written equivalent and direct download.
Isolated AI-assisted delivery
A real worktree, tmux, Neovim and authenticated AI delivery loop applied to a synthetic production-policy defect.
A real worktree, tmux, Neovim and authenticated AI delivery loop applied to a synthetic production-policy defect.
Written walkthrough
- Read the synthetic brief, run the sandboxed baseline suite and confirm a clean main branch.
- Create an isolated worktree through the real tmux and Worktrunk workflow, then inspect its dedicated ports.
- Open Neovim and authenticated ocv side by side; ask the agent to identify the exact two-file boundary.
- Add the behavioural test by hand and prove the expected red state before allowing one implementation edit.
- Run the sandboxed suite, review the bounded diff, selectively stage both files and verify a clean committed handoff.
Designed outcome: The agent accelerates a bounded change while tests, scope gates and human review remain the approval boundary.
Kubernetes incident response
A repeatable inspect, contain, change and verify sequence against a synthetic cluster fixture.
A repeatable inspect, contain, change and verify sequence against a synthetic cluster fixture.
Written walkthrough
- Inspect workloads in the synthetic storefront namespace.
- Identify the simulated API pod in CrashLoopBackOff.
- Read previous-container logs and locate the missing configuration.
- Review the fixture manifest diff before applying it.
- Confirm the simulated rollout returns both pods to Running.
Designed outcome: The responder follows evidence instead of improvising against production.
Vim and tmux delivery workspace
A focused edit, test and review loop that keeps the working set visible and keyboard-addressable.
A focused edit, test and review loop that keeps the working set visible and keyboard-addressable.
Written walkthrough
- Open a synthetic refactor workspace in a named tmux session.
- Place implementation and tests side by side in Vim.
- Make a bounded parser change and write both fixture files.
- Switch to the checks window and run deterministic tests.
- Return to the editor and inspect the compact diff summary.
Designed outcome: The terminal becomes a reliable workspace, not a collection of memorised tricks.
Reproducible environment bootstrap
A pinned, diagnostic-first workstation bootstrap using synthetic manifests and fixture paths.
A pinned, diagnostic-first workstation bootstrap using synthetic manifests and fixture paths.
Written walkthrough
- Read the pinned synthetic toolchain manifest.
- Verify platform support and manifest validity before installation.
- Simulate shell, editor and runtime setup.
- Dry-run configuration links and report conflicts before applying them.
- Run path, version and configuration diagnostics against a sample app.
Designed outcome: A new environment is explainable, repeatable and recoverable.
Start with the constraint
Bring the workflow that keeps breaking.
On the booking form, choose the area you want to improve. We will use the first call to decide whether this needs a team audit, a focused implementation or simply an honest answer.
Book a thirty-minute intro