How I work with AI.

I use AI to explore quickly, then slow down to decide carefully. It compresses the parts of design where I'd otherwise spin, but the framing, the empathy, and the final call stay mine.

Agents I've built

04 / experiments + production

Live ·Personal ·2025–present

Job Hunt Master

An automated job-hunting agent that runs on a schedule. Pulls from job boards, scores fits, emails me curated matches each morning, and keeps a running tracker of where I've applied.

Claude·Scheduled tasks

Live ·Personal ·2025

Research Synthesis Bot

Takes 20+ interview transcripts and surfaces themes, edge cases, and verbatim quotes I would've missed. Claude Project with a custom prompt library.

Claude Projects·Prompt library

Live ·SubjectToClimate ·2025–present

Synthetic Persona Generator

Generates research-grade synthetic personas so the team can stress-test product concepts before going to real teachers and students. Built in my Agentic User Researcher role at SubjectToClimate.

Claude·Custom prompts

Live ·Personal ·2025

Design Critique Buddy

Reviews my Figma flows against heuristics and accessibility guidelines. I describe the screen, it pushes back with the questions a senior would ask.

Custom GPT·Heuristics

I believe

Good AI use in design is about compressing exploration, not compressing judgment. The faster I can play with options, the more time I have for the part only I can do, talking to real people, framing the right problem, and choosing what ships.

My AI stack

what I actually open every week

Claude Synthesis, writing, sparring

Figma Make AI-assisted design exploration

Cursor Vibe-coding prototypes

Perplexity Domain research with sources

ChatGPT Quick inference, voice mode

Gemini Long-context analysis

NotebookLM Long-doc synthesis

AI in my design process

where it actually fits

Where I draw the line

what stays mine

✓ What I let AI do

Pattern-finding across messy input Surfacing themes from 30+ transcripts before I synthesize manually.
Drafting + pressure-testing copy First drafts, plus playing skeptical user / busy exec / new joiner.
Quick prototypes that test a feeling Sketch to clickable in a morning, so the team can react to something real.
Strategy & framework drafting Competitive teardowns, decision matrices, opportunity sizing, first pass.
Heuristic + accessibility checks A second pair of eyes that never gets tired before I do.

× What stays mine

Problem framing The instinct for what's actually worth solving comes from human conversation.
User empathy Real interviews, real silence, real body language. AI summarizes; it doesn't witness.
The final call on what ships Tradeoffs are about taste and stakes, not optimization.
Sensitive personal data Anything user-identifiable stays out of cloud models, full stop.
Original creative direction Visual taste is a muscle, outsourcing it makes it weaker.
Anything I'd be embarrassed to defend If I can't explain why something shipped, AI didn't help, it covered for me.