Agentic Product & Automation Engineer

Brent Stees

Engineering leader scaling web products through creative direction and agentic automation.

Read this page as

Strategy, teams, and creative direction.Architecture, automation stack, and evals.

Color theme

01 · About

I've spent my career leading software teams and directing the craft of what they ship. I've designed, built, and shipped hundreds of sites, and I keep making complex systems clearer for the people who build and use them.

I'm an engineering manager who retooled as a hands-on agentic product and automation engineer. Today I build the agents, pipelines, and interfaces that let web apps scale, and I still write the CSS myself.

Lately the work is about operating models for human-and-agent teams: where AI accelerates delivery, where humans must decide, and how to make that boundary explicit, reviewable, and repeatable.

My public work leans on explicit contracts: MCP tools, versioned asset sync, deterministic indexes, and evaluation harnesses for agent instructions. If an agent workflow can't be observed and tested, it isn't finished.

  1. 01

    Context before prescription

    Understand the system, its constraints, and the people inside it before choosing the mechanism.

  2. 02

    Make the right path easier

    Good enablement turns quality into the path of least resistance, not another checklist.

  3. 03

    Feedback is a product surface

    The speed and clarity of learning shape the quality of every decision that follows.

02 · Selected work

01 · Engineering operating system

AI Native

The central standards, workflows, skills, and governance layer for a hybrid human-and-agent development model. Product work happens in product repos; this repo improves how those repos work.

Why it mattersMakes AI-assisted delivery repeatable without removing human judgment. Teams inherit a shared operating surface, decisions are written down with their reasoning, and risky work routes through explicit human review.

How it worksA versioned asset pipeline: shared standards and skills sync into consumer repos as a managed overlay, with drift triage, calver releases, a deterministic repo-surface index, and an instruction-evaluation harness for AI-facing Markdown.

System map: select a node

Shared assets Standards, workflows, skills, sub-agent profiles, context policy, routing, and provider adapters — the operating surface every repo inherits.

  • Node
  • React
  • Vite
  • Express
  • D3
  • Docker
  • MySQL

02 · AI workstream orchestration

Orca

Track many parallel AI coding sessions from one dashboard, and get a push notification the moment one needs a decision.

Why it mattersParallel agents only scale if the human stays the bottleneck by choice, not by accident. Orca keeps every session legible and turns “is it blocked?” into a notification instead of a polling habit.

How it worksA Fastify hub exposing MCP tools and a WebSocket feed. A VS Code extension registers each window and reports idle state; sessions call orca_request_input, which raises a dashboard banner and an ntfy push. State lives in SQLite; an Electron shell bundles server and UI.

System map: select a node

Orca server The central hub on :7337 — tracks sessions, serves the dashboard, and exposes MCP tools over a WebSocket channel.

  • MCP
  • Fastify
  • WebSockets
  • React
  • VS Code API
  • SQLite
  • Electron

03 · Capabilities

  1. 01

    Agentic workflows

    Operating models, governance, and human decision gates that make AI-assisted delivery trustworthy at team scale.

    Multi-agent orchestration, MCP servers and tools, context policy, routing, and evaluation harnesses for prompts and instructions.

  2. 02

    CSS & UX architecture

    Hundreds of sites designed, built, and shipped. Clear hierarchy, reusable patterns, and critical flows treated as one system.

    Token-driven design systems, fluid type, container queries, :has()-driven state, and motion that respects reduced-motion users.

  3. 03

    Creative direction

    Setting the bar for craft across a team — a point of view on quality that people can recognize and repeat.

    Turning a direction into constraints: type scales, spacing systems, component contracts, and review criteria.

  4. 04

    Scaling web apps

    Leading engineering teams through growth — paved paths, onboarding, and standards that make the right work easier.

    CI/CD, repeatable environments, sync and rollout pipelines, and automation that removes avoidable toil.

  5. 05

    Observability & feedback

    Shortening the distance between intent, implementation, and evidence so decisions happen sooner.

    Session state, health signals, review queues, live dashboards, and fast test harnesses that keep work legible.

04 · Contact

Have a complex system to make clearer, or a workflow worth automating?

I'm open to conversations about agentic product engineering, AI-native development, developer experience, and hands-on product systems work.

Connect on LinkedIn