• Product design
  • System design
  • AI as system layer

I build systems that
make complex things usable.

I design and build product systems end to end. Most of the work lives inside long-running platforms where structure and consistency actually matter, and the design problem is usually the constraint, not the canvas.

More recently, that work has stretched into AI-native products, where the model is one constrained stage in a structured pipeline and prompt engineering is part of the system, not the chat.

See the work Get in touch
How I work

AI as a system layer.

I don't use AI as a tool I jump into. I use it as a structured layer between intent and execution.

Every project starts with rough specs. Before anything is built, I use LLMs to interrogate the problem, forcing clarity on requirements, constraints, and edge cases. The model asks the questions a strong engineer or product partner would ask.

From there, I translate that into structured prompts that are scoped, reusable, and tied to the system they operate in.

Execution isn't a pipeline. It's environment selection:

  • Claude — agentic work (reasoning, generation, code)
  • Cursor — local iteration (code, refactor, integration)
  • Claude Design — visual exploration, when needed

Claude and Cursor both handle reasoning and code. The choice is contextual, not by role. Claude Design enters only when visual reasoning moves the work forward faster than code.

The model is one constrained stage. The system is what matters.

More about how I work
From intent to ship (AI-native system)
From intent to ship · AI-native system
Intent Interrogate Structure (Prompt System) Execute Claude agentic work — reasoning, generation, code Cursor local iteration — code, refactor, integration Claude Design visual exploration, when needed Review Ship

Execution environment is selected per task, not by role.

Build

Things I've built,
and systems I work in.

A flagship AI product on one side, a multi-site platform and financial advisor systems on the other. All of them have to keep running while they get better.

Selected projects Things I've built.
Flagship
Flagship

ChatOBD2

AI-native automotive diagnostics interface

  • AI-native UX
  • System translation
  • Technical interaction design
Product

Sawa Marketing Site

Marketing site and content surface for a group-decision app

  • Marketing site
  • Content structure
  • Frontend implementation
Experience Systems I operate within.
Platform Work · MagTek

MagTek Platform

UX systems across a multi-site enterprise platform

  • Platform-level UX
  • Frontend systems
  • System evolution under constraints
Product Work · InvestCloud

Advisor Platforms

Advisor-facing platforms across major wealth institutions

  • Multi-tenant frontend
  • Structured layouts for dense data
  • Per-institution adaptation
Design workflow

Claude Design (visual branch).

Claude Design is one branch of the execution layer. I use it when structure alone isn't enough and the problem needs to be seen.

That usually means:

  • new flows
  • layout and hierarchy decisions
  • comparing multiple directions quickly

It's not part of every loop. It earns its place when visual reasoning moves the work forward faster than code. Once direction is clear, everything resolves back into the same system and build loop.

More about the design workflow
The visual branch of the execution layer
Visual branch
Visual intent Prompt Claude Design Codebase

One branch of the execution layer. Selected when the work is easier to reason about visually than in code.

Get in touch

If you're building something interesting,
drop a note.

A few words is plenty. Replies usually within a day.

No tracking. Replies usually within a day.

Want more space? Use the contact page.