What I work on, in detail.

What you get

• Architecture diagram and a written rationale for each non-obvious choice.
• Data model with explicit role and permission semantics.
• API contracts (OpenAPI / GraphQL schema / typed RPC).
• Audit-log strategy and threat model when the data is sensitive.

When you'd reach for this

• Multi-tenant or multi-role product where access matters.
• Replacing a legacy system that grew organically.
• Pre-launch sanity check on someone else's design.

What you get

• A production web application — frontend, backend, database, deploy.
• CI/CD pipeline that catches regressions before they ship.
• A test suite scoped to what actually breaks (not coverage theatre).
• Documentation and runbooks aimed at the team that inherits the code.

When you'd reach for this

• You need a real web app and don't have a senior developer in-house.
• Existing build is slowing your team down and needs a clean rewrite.
• Greenfield product where you want one set of hands across the stack.

What you get

• A working tool deployed on day one of week one — no big-bang launch.
• Role-aware views (operator, supervisor, admin) from the same codebase.
• Filters, sorts, exports, bulk actions — the boring features that matter.
• Reports that go to email, Slack, or wherever the people who need them already are.

When you'd reach for this

• Your team is doing the same manual workaround in spreadsheets every week.
• Customer support is asking engineering for one-off database lookups.
• You're hiring more ops staff to cover what software should be doing.

What you get

• Honest written assessment of where LLMs help and where they're a distraction.
• Prototype features (RAG, structured output, classification, agents) you can put in front of users.
• Evaluation pipeline — so you know if quality regresses with the next model swap.
• Provider strategy — when to pin to a vendor, when to stay portable.

When you'd reach for this

• Your team wants to ship AI features but you're not sure which ones.
• You've shipped an LLM feature and don't trust the output.
• Board asking about AI strategy and you want a real answer.

What you get

• Reliable integrations that handle retries, idempotency, and rate limits.
• Background workers and scheduled jobs with monitoring.
• A clear, written map of every integration — what it does, what breaks if it stops.
• Alerting that pages the right person, not the whole team.

When you'd reach for this

• Data is being copy-pasted between two systems by a human, daily.
• A vendor migration is coming and you need a clean handoff.
• A scheduled job has been "mostly working" for years and finally broke.

What you get

• Codebase audit with prioritized findings (security, performance, fragility, debt).
• Monthly retainer covering bug fixes, small features, and operational support.
• Gradual modernization — incremental rewrites that never require a "big rewrite."
• On-call rotation that respects your team's actual coverage needs.

When you'd reach for this

• Software you depend on is older than its original developer's tenure.
• Original team has moved on and the docs disappeared with them.
• You want senior coverage but don't need a full-time hire.

What you get

• Written architecture review with specific, prioritized recommendations.
• Code review of a feature, a refactor, or a whole repo.
• Hire decision support — interviewing for senior engineering or technical leadership.
• Vendor evaluations — "should we buy X or build it ourselves?"

When you'd reach for this

• You're about to commit to a big architectural decision and want a sanity check.
• You're hiring senior engineering and want a technical interviewer.
• A vendor pitch sounds too good and you want it stress-tested.