ndrew weyrich putting the pieces together

AGENTIC EXPERIENCE DESIGN

Deploying AI Agents that super charge my ideation process by picking the best frontier LLM for each design task to create full prototypes and ship code.

Currently, Lead UI UX Product Designer at Peraton.

Agents · Frontier LLMs · Prototypes · Shipped Code

Model Routing Agent Workflow Get in touch

Model Routing

The right frontier LLM for every design task.

No single model is best at everything. I route each step of the design process to the model that does it best, then let agents hand work off between them.

01

Ideation & research

Agents gather market signals, competitor patterns, and user pain points, then turn them into product directions worth testing.

  • Perplexity Computer — research & orchestration
  • ChatGPT — concept generation
  • Grok — real-time trend signals
02

UX writing & flows

Long-context reasoning models map user journeys, edge cases, and microcopy so every screen has a purpose before pixels are drawn.

  • Claude — flows & UX copy
  • ChatGPT — naming & positioning
  • Gemini — accessibility review
03

Visual & prototype

Multimodal models critique layouts and generate interactive prototypes that feel like the real product on day one.

  • Figma Make — interactive UI prototypes
  • Gemini — screenshot & visual critique
  • Claude — front-end prototype code
04

Build & ship

Coding agents turn validated prototypes into production code, with tests, reviews, and a clean handoff to engineering.

  • Claude Code — production code
  • Codex — refactors & tests
  • Cursor — in-editor pairing

Agent Workflow

From first idea to shipped code.

A repeatable agentic loop that compresses weeks of ideation, prototyping, and build into days — with a designer directing every decision.

01

Frame the problem

I define the user, the outcome, and the constraints, then brief the orchestrating agent with a clear design goal.

  • Problem framing
  • Success metrics
  • Agent brief
02

Route to the best model

The orchestrator assigns each subtask to the strongest frontier LLM and runs them in parallel.

  • Model selection
  • Parallel agents
  • Cross-model review
03

Prototype end to end

Agents assemble full, clickable prototypes — real data, real flows, real interactions — ready for user testing.

  • Clickable prototypes
  • Usability testing
  • Rapid iteration
04

Ship and learn

Coding agents push production-ready code, and analytics feed the next round of ideation.

  • Production code
  • Deploy & monitor
  • Analytics loop

Get in touch

Do you need an Agentic Experience Designer?

I orchestrate AI agents across frontier LLMs to take ideas from first sketch to working prototype and shipped code.