Agentic engineering setup
I've shipped production code with AI coding agents since 2024. I set your team up to do the same — faster, and without the mess.
- —Make the codebase agent-ready: conventions, guardrails, tests and CI that let agents move fast without breaking things
- —Design the workflow — what agents draft, what humans review, how work is specced, verified and merged
- —Coach engineers on effective AI-assisted development; at Sapien the practice cut one project's build time by 50%
- —Pick and wire the tooling — coding agents, MCP servers, review bots — and measure whether it's actually paying off
- Claude Code
- Codex
- MCP
- GitHub Actions
real numbersProduct delivery, web & mobile
The whole product, from Figma to production — one engineer, an agentic workflow, and a scope that would have taken a team.
- —Web apps in Next.js/React that load instantly, rank well and turn visitors into customers
- —Mobile apps in React Native for iOS and Android from one codebase — including store submission and the yearly OS breakages
- —Cost-optimised hosting that scales with traffic — typically 40–60% cheaper than what it replaces
- —Rescue and speed up existing apps — code, caching and infrastructure fixes, often 2–5× faster loads
- Next.js
- React Native
- TypeScript
- Vercel
AI features in production
LLM features and agents that survive contact with real users — with evals, observability and a bill you can predict.
- —Build tool-using agents and LLM features on your data and your stack (Claude, OpenAI, open models)
- —Set up evals and observability so quality is measured, regressions are caught and prompts can change safely
- —Keep cost and latency under control: caching, model routing, batching, budgets
- —Handle safety and data: guardrails, permissions, PII, audit trails — before it becomes a support ticket
- Claude API
- OpenAI
- pgvector
- evals
APIs, cloud & platform
The boring part that makes the fast part safe.
- —REST or GraphQL APIs that are secure, fast, and documented from the code so the docs stay true
- —Event-driven and serverless back-ends that react in real time and cost nothing when idle
- —CI/CD and infrastructure-as-code: deploys in minutes, environments reproducible and versioned
- —Cloud architecture that absorbs traffic spikes, stays available and doesn't surprise you on the invoice
Fractional CTO & technical leadership
A senior engineer in the room when the decisions get made — for a day a week or a critical quarter.
- —Roadmap and architecture: what to build, what to buy, what to postpone, and why
- —Migrations everyone's afraid of: platform, framework, cloud — planned, de-risked and shipped
- —Hiring and mentoring engineers, and setting the practices a small team needs to stay fast
- —Due diligence and audits of an existing codebase, team or vendor, in plain language
- 1 day / week
- one quarter
- one-off audit