How i build with AI

Custom Agents: I use AI to build custom internal agents that connect research, documentation, roadmaps, and design systems into a unified workflow layer with Claude, ChatGPT, Perplexity, and Copilot. These agents are tailored to specific product and business contexts, enabling teams to produce more consistent, scalable outputs while reducing friction across the design and development process.

Features & Experiences: I also use structured prompting systems, specifically the RTCCF framework, to translate internal design systems and product context into meaningful, repeatable design iterations. Through this approach, I was able to build an entire functional app through vibe coding, including SquareUp, a restaurant dashboard and check-splitting app. This system combines AI-assisted workflows with tools like Claude, Replit, Builder.io, and Figma Make.

Custom Agents

TripAdvisor AI Agent “ollie”

The custom AI agent shown in the video was designed to support scalable cross-functional execution by consolidating research, product planning, and system dependencies into a single workflow. It was trained on internal UX research, product roadmaps (via Jira and Airtable), and the TripAdvisor design system, ensuring outputs stay grounded in real product constraints.

The agent helps teams proactively plan feature work by connecting what is being built to what already exists in the system, reducing ambiguity between design, product, and engineering and improving alignment across execution.

Where It Adds Value

Research Synthesis — Aggregates UX findings and customer insights into structured outputs

Roadmap Visibility — Surfaces project timelines, priorities, and dependencies across teams

Component Impact Mapping — Identifies affected design system components and templates

Documentation Intelligence — References existing product and design system documentation

Cross-Team Alignment — Improves collaboration between design, product, and engineering

Operational Efficiency — Reduces manual coordination and repetitive workflow overhead

Scalable Product Execution — Leverages existing data and systems to accelerate feature delivery

BUILDING Features & Experiences with AI

The prompt framework

I use a framework called RTCCF to create more reliable AI generations across product design, UX workflows, and full application experiences. Using this approach, I built a real working application called SquareUp for a restaurant to streamline the guest checkout process through QR-based payments while also giving management real-time visibility into nightly sales performance and operational insights. The system improves table turnover, reduces server workload, and provides actionable data that helps teams understand how the business is performing in real time.

  • Role — Defines the expertise and perspective AI should operate from

  • Task — Establishes the exact objective or workflow outcome

  • Context — Provides product, user, and business constraints

  • Constraints — Controls quality, structure, accessibility, and system rules

  • Format — Shapes how outputs are generated and organized

The image on the right shows my actual implementation of the RTCCF prompt format for this app at a high level.

The Product

SquareUp was built as an exploration into how structured AI prompting and product context can accelerate production-ready product design workflows. The project reinforced a key principle when working with AI systems: context is everything.

Creating usable AI outputs requires far more than a simple prompt. To generate scalable, developer-ready experiences, prompts must be intentionally structured around real UX and UI foundations.

UI Context

AI outputs were guided using structured systems that defined:

  • Grid systems and spacing rules

  • Typography, color tokens, and design system constraints

  • Responsive behavior across screen sizes

  • Scalable component architecture

  • Accessibility and usability best practices

UX Context

Prompts were also tailored around user behavior and product strategy, including:

  • End-to-end user flows and journeys

  • User intent and decision points

  • Interaction patterns and feature logic

  • Research-informed UX frameworks

  • Clear action hierarchy and navigation structure

Workflow & Execution

Using structured prompting frameworks like RTCCF alongside tools such as Claude, Replit, Builder.io, and Figma Make, I was able to build SquareUp, a restaurant dashboard and check-splitting application.

These systems helped generate outputs that could move beyond ideation into production-oriented execution by:

  • Accelerating design iteration and feature exploration

  • Producing cleaner, more implementation-ready UI

  • Connecting outputs into Figma through MCP workflows

  • Supporting developer-ready design handoff

  • Reducing repetitive production overhead across workflows

The pitch deck

SquareUp started with a clear problem of simplifying restaurant operations and split payments into a faster and more intuitive experience. That opportunity was translated into structured user flows and a prompting framework grounded in UX principles.

Using structured prompts, I defined journeys, interactions, and UI logic before generating features. The result was a set of production-ready outputs including a restaurant dashboard, check-splitting flow, and intuitive payment experience.

View slides