01 / 04
Capabilities · Engagements
What I build, and how I build it.
Three categories of work that span the practice — built on an AI-native workflow that lets one senior engineer move at the speed and breadth of a team. I'd rather take on five engagements a year and ship them than thirty and miss. Each pillar describes the kind of project I take on, the stack I'd use, and the shape of problems that fit.
Recently shippedSquadUp (live on the App Store) · YardPaint (consumer AI app) — both designed, built, and shipped end to end.
See the studio →Pillar / 01 of 03
01
Full-stack applications
Greenfield builds and rebuilds, end to end.
Most of what I do is end-to-end web applications — customer portals, internal tools, operational dashboards, customer-facing products. On greenfield projects I default to TypeScript and Postgres — well-understood tools that hold up over years. On rebuilds and joins to existing systems, I work in whatever stack you're already running. I write the code myself.
Spec
- Default stack
- Next.js · TypeScript · Postgres · Tailwind
- Comfortable in
- React · Vue · Python · Go · Rails · MySQL · MSSQL · MongoDB
- Native
- Swift / SwiftUI when iOS is in scope
- Hosting / runtime
- Cloudflare · Vercel · Fly.io · AWS · Azure · GCP · Docker · Kubernetes
- Typical scope
- 6–14 weeks · fixed price or T&M
- Team size
- 1 engineer (me)
Problems that fit
- Member portals and customer accounts that need to handle real volume
- Internal tools replacing the spreadsheet-and-Zapier sprawl
- Billing, admin, and audit consoles wired to Stripe or your ERP
Pillar / 02 of 03
02
Cloud, pipelines & automation
The infrastructure and pipelines that move your data from systems to decisions.
Cloud infrastructure and CI/CD — containerized, observable, and reproducible. Data pipelines that turn scattered operational data into live dashboards and reports, and then into automated decisions, with AI in the loop where it pays off. The deterministic plumbing that sits under the AI layer and keeps it honest.
Spec
- Cloud
- AWS · Azure · GCP · Cloudflare
- Containers / orchestration
- Docker · Kubernetes
- CI/CD
- GitHub Actions · pipelines · infrastructure as code
- Data / pipelines
- Python · SQL · Airflow · Temporal · scheduled jobs
- Surfacing
- Dashboards · reports · alerts → decision automation
- Typical scope
- 3–8 weeks
Problems that fit
- Standing up cloud infra and CI/CD that a small team can actually operate
- Turning scattered operational data into dashboards and reports people trust
- Replacing brittle manual steps with observable, recoverable automated pipelines
Pillar / 03 of 03
03
AI systems & agents
AI-augmented systems that do operational work, with the rigor to run unsupervised.
Production agents and pipelines built on Claude, the Claude Agent SDK, MCP, and OpenClaw when an open-source framework fits the deployment shape. The difference between an agent demo and an agent in production is observability, audit trails, and the ability to roll back when one gets a step wrong. I build the version you can leave running.
Spec
- Models
- Claude (Opus, Sonnet) · OpenAI when warranted
- Frameworks
- Claude Agent SDK · OpenClaw · MCP servers · custom tool harnesses
- Runtime
- Cloud Run · Workers · Kubernetes · Docker · self-hosted runners
- Observability
- Langfuse · OpenTelemetry · custom
- Pipelines
- AI evaluation in CI · prompt regression suites · model deployment
- Typical scope
- 4–10 weeks · fixed price
Problems that fit
- Bug-to-PR engineering agents that take a flagged exception, reproduce it, and open a draft pull request with passing tests
- Computer-use agents that drive workflows locked behind legacy UIs without APIs — the operational work that's been too brittle to automate before
- Multi-agent pipelines that own a multi-step operational process end to end, with audit, rollback, and human-approval gates where they matter
What I don’t do
Being clear about what falls outside the practice makes the rest of the conversation easier.
- Strategy decksI write code, not strategy decks. If you need a deck, I'll point you to someone who does that well.
- Pure design workI ship usable interfaces. For visual design as the primary deliverable, you want a design studio.
- Offshore-rate competitionPricing reflects senior US-based engineering. That's the trade.
- Staff augmentationI scope and own deliverables. Engagements where I'm a long-term seat in someone else's team aren't the model.
- Long vendor-onboarding cyclesI work in weeks, not quarters. Multi-month security reviews don't fit the engagement size.
- Productized SaaS (today)I ship custom builds. Products are a side effort.
Next step