02 / Practices

A short list we can actually staff.

We do not sell “digital transformation” or last decade’s process. We sell an intelligent crew that sits in your repo, uses AI as a tool they can defend, and finishes the work.

Product engineering

Embedded teams for products that have left the prototype and not yet found a durable shape. We do not start with a 40-page spec. We write the features, the tests, and the decisions we declined — with AI in the loop, not as a ceremony.

0 → 1 products that need a second architecture
Internal tools that became the actual product
Migrations that cannot take the site down

TypeScript · Go · Python · Rails · React

Platform & runtime

Infrastructure as a product for the people who ship on it. Fewer tickets to platform. Faster path from merge to production. Clearer failure modes.

Deploy paths and environment parity
Observability that on-call can read
Data stores, queues, and the jobs that feed them

Kubernetes · Terraform · Postgres · AWS

Interface systems

We design for operators, not shoppers. Density, keyboard, empty states, and error copy are the work — not the garnish.

Admin surfaces and command palettes
CLI + dashboard noun alignment
Design systems that live next to application code

UI · DX · Design tokens

Applied AI

Models are cheap. Product judgment is not. We put agents and retrieval inside existing workflows — and we use the same class of tools to build faster than a shop still running last decade’s methodology. Then we measure whether anyone uses it next week.

Tool-use and MCP surfaces for real products
Eval harnesses before the launch email
AI-assisted build, human review on the decisions that matter

Python · TypeScript · Evals · Agents

Retainers

Most products need a studio on call after the first ship. A small, named crew. Same repo. No bait-and-switch bench.

Monthly capacity after a build
Incident pairing and hardening
The same people who shipped it, still answering

Your stack · your hours