Five ways in.
One outcome.
AI-built MVPs arrive fast, fragile, and unsafe. Every engagement here ends in production-grade: fixed price, async-first, delivered by the engineer accountable for the code.
The $250 48-hour Teardown.
Zero-risk guarantee5+ production risks you didn't know about, or you don't pay.
/ what it is
Send your live URL. In 48 hours I open the repo, trace the architecture, profile the queries, check the auth, and flag every API the AI hallucinated and every edge case it skipped. You get a recorded walkthrough and a one-page report naming the risks that matter.
/ what I deliver
- One-page risk report — the five-plus production risks ranked by severity, readable by non-engineers.
- Recorded Loom walkthrough — key findings explained on screen, shared within 48 hours of start.
- Priority sequence — what to fix first, what to defer, what to ignore.
- Fixed-price remediation quote — if you want me to fix what I found. Credited to the Architecture Sprint.
/ what I check
- Security — auth flows, RLS enforcement, input validation, secrets handling.
- Architecture — component coupling, state management, module boundaries, AI-generated code patterns.
- Fragility — hallucinated APIs, missing edge cases, inconsistent error handling.
- Performance — query efficiency, bundle weight, slow paths under concurrent load.
/ who it is for
- Founders with an AI-built MVP who want to know what breaks before users do.
- Teams preparing for a launch or funding announcement.
- Anyone who inherited AI-generated code and needs a second opinion before touching it.
If you already know exactly what's broken and just need someone to fix it, book the Architecture Sprint or Production Build directly. The Teardown pays off when the answer isn't obvious.
Architecture Sprint.
/ what it is
Before any production code, the build gets de-risked. In 48 hours: a clickable prototype, a database schema, and a technical blueprint with a fixed-price quote. You see what you're buying before you commit.
Guarantee: a blueprint any engineer could ship from, or a full refund. Credited in full to the Build if you sign within 14 days./ what I deliver
- Clickable prototype — a working UI skeleton showing navigation and core flows.
- Database schema — PostgreSQL tables, relationships, RLS policies, and migration plan.
- Technical blueprint — stack decisions, API contracts, third-party integrations, auth model.
- Fixed-price build quote — a milestone-based quote any engineer could execute from.
- 15-min Loom walkthrough — every decision explained on screen.
/ how it connects
- Typically follows a $250 Teardown when deeper scoping is needed.
- $900 is credited in full to the Production Build if you sign within 14 days of delivery.
- The blueprint is yours to keep regardless of what you decide next.
/ what I need from you
- A brief describing the core problem the product solves.
- Any existing design files, user research, or prior technical decisions.
- Access to the repo or any AI-generated codebase I'm building on top of.
- One stakeholder who can define done in measurable terms.
If your architecture is already decided and you just need execution, skip to the Production Build. The Sprint is for founders who need clarity before committing to a full build.
Production AI workflows.
/ what it is
Not generic chatbots. Specific AI systems that hold up on real data: customer support that triages itself, documents that process themselves, knowledge bases your team queries in plain English. Integrated into your infrastructure, guardrailed, production-safe.
You get an AI system that holds up under real users and real data, not a demo that breaks the first week./ what I deliver
- Agent scoping document — input/output spec, success criteria, cost model.
- Technical architecture blueprint — LLM selection, pipeline design, data flow.
- Production agent code — deployed, typed, tested on critical paths, with guardrails and fallbacks.
- Prompt library — all system prompts, versioned, documented.
- Monitoring setup — logging, cost tracking, error alerting.
- 30-day post-launch support — bug fixes and prompt tuning included.
/ complexity tiers
- Simple (single LLM call, structured output) — 1–2 weeks, $6,000–$8,000.
- Standard (multi-step pipeline, RAG, tool use) — 3–4 weeks, $14,000–$20,000.
- Complex (multi-agent orchestration, production UI) — 5–6 weeks, $22,000–$28,000.
/ what I need from you
- A clear problem statement: what manual process are we replacing?
- Sample data or documents the agent will process.
- Access to existing systems the agent needs to integrate with.
- One stakeholder who can define success in measurable terms.
If you want a chatbot that answers generic questions from a PDF, use an off-the-shelf tool. I build custom AI systems for specific business workflows where reliability and integration matter.
Production Build.
/ what it is
For seed-stage founders who need to show investors a working product, not a pitch deck. From an AI-built demo or greenfield brief to a production-grade V1: auth, payments, database, core business logic, and responsive UI. Typed, tested on critical paths, RLS-secured. Shipped in weeks, not months.
Agencies quote 10–12 weeks and $60k for this. It ships here in weeks, from the engineer accountable for the code./ what I deliver
- Up to 5 dynamic views, fully responsive, with real data.
- Auth and payments — Supabase Auth, Stripe checkout, subscription logic.
- Database and API — PostgreSQL schema, RLS policies, server actions.
- Testing — typed, tested on critical paths (Vitest), 80% coverage target.
- Deployment — production-ready on Vercel, codebase walkthrough Loom.
- 30-day post-launch support — bug fixes and critical patches included.
/ how it starts
- $250 Teardown or $900 Architecture Sprint first. I don't write production code without understanding the problem.
- Contract and 40% upfront — MSA and SOW signed before the first line of code.
- Daily Loom updates — progress visible every day, staging link updated daily.
- 30% at mid-milestone, 30% on delivery — IP transfers to you on final payment.
/ the stack
- Next.js 15 — App Router, TypeScript strict, React Server Components.
- Supabase — PostgreSQL, Auth, RLS, Edge Functions.
- Stripe — checkout, subscriptions, billing portal.
- Vercel — deployment, preview URLs, edge network.
- PostHog — analytics. Resend — transactional email.
If you need a landing page, use Framer. If you need a simple internal tool, use Retool. I build custom production applications where architecture, security, and scale matter from day one.
Reliability retainer.
/ what it is
For teams that shipped a V1 and need an embedded engineer for the next 90 days. Features, bug fixes, migrations, performance tuning, and architecture guidance, delivered weekly and async with direct Slack access. The codebase stays production-grade instead of drifting back to fragile.
/ what you get
- Weekly feature releases — 1–3 features per week, prioritized via shared Linear board.
- Bug fixes — critical within 24 hours, non-critical within the sprint.
- Architecture reviews — bi-weekly 15-min Loom on codebase health.
- Direct Slack access — for technical questions and architecture advice.
- Monthly production health review — query optimization, dependency updates, security patches.
/ what it isn't
- Not a full-time seat or a fractional CTO arrangement.
- Not 24/7 on-call. This is a deliberately lean practice; the rate reflects that.
- Not ad-hoc meetings. Async-first, Loom-driven, daily staging updates.
/ how to start
- Production Build or Audit first. The retainer makes sense after I know your codebase.
- Three-month minimum, then month-to-month with 30-day notice.
- Cancel any time after the first three months. No exit theatrics.
If you need someone to ship features 40 hours a week, hire a full-time engineer. The retainer is for maintaining velocity and architecture quality, not replacing an engineering team.