§ 03

AI Work

Shipped,
not specced.

There's a version of "AI experience" that exists entirely in interviews. This isn't that. Everything here is in production or built end-to-end as a functioning application.

§ 03.1

Production · Exclusive Resorts

AI shipped at scale
in a luxury travel platform.

Some of these are member-facing, some are internal tools for advisor teams — both add real operational and commercial value running on production traffic.

What we've shipped

Trip Recommendation Tool Natural language trip planning assistant — members describe what they're after, the tool surfaces matched destinations and itineraries from inventory, drawing from CMS content and concierge-vetted options first
Trip Itinerary Creator Generates full activity and dining itineraries based on known member preferences and a few quick questions — prioritizes CMS and concierge-vetted content, supplements with broader sources when needed
Member Profile Summarization LLM-generated profile summaries giving advisor teams instant context on member preferences, travel history, and patterns — reducing ramp time before every interaction
AI Chat Agent Conversational agent handling member inquiries with access to property data, availability, and member records
LLMsRecommendation SystemsRAGAWS BedrockSalesforceBrazeAlgolia

§ 03.2

Personal Project · Built in Bolt

CrewChief — AI-powered
automotive ownership platform.

Full-stack web application I built to solve a real problem: most vehicle owners have no organized system for their car's knowledge, history, and future work. CrewChief fixes that with a persistent AI advisor that actually knows your car.

Core concept

When you add a vehicle, the app calls Gemini to generate a comprehensive dossier — known failure points with mileage ranges, full factory maintenance schedule, fluid specs, common modifications, reliability score, and powertrain details. Everything downstream builds from that knowledge base.

What's under the hood

AI Consultant ("CrewChief") Gemini 2.5 Pro chat initialized with a 200+ line system prompt injecting the full vehicle record — mileage, service history, tracked issues, mods, recalls, and more.
Agentic Write-Back The consultant emits structured command tags mid-conversation that the backend parses and executes — adding wishlist items, closing issues, updating mod status — without the user touching a separate UI.
Invoice OCR & History Upload a service invoice image; Gemini 2.0 Flash extracts date, cost, vendor, mileage, and every line item into a timestamped maintenance history the AI references in future sessions.
Cost Estimation Engine Request a shop quote for any wishlist item — itemized parts + labor ranges with regional rate acknowledgment, validated against a strict Zod schema.
Labor Bundling Optimizer Each service item is assigned a physical location zone by the AI. The bundler flags items sharing access requirements and estimates labor savings from doing them together.
Model Tiering by Task Three Gemini models allocated by cost/capability: 2.5 Flash (structured JSON, temp 0.3); 2.5 Pro (temp 0.7) for conversation; 2.0 Flash for invoice vision/OCR.
Next.js 13TypeScriptSupabaseGemini APIZodshadcn/ui
Live app →
crewchief-gemini-aut-fa6t.bolt.host

CrewChief

AI-powered automotive ownership platform — persistent vehicle knowledge base, agentic write-back, invoice OCR, cost estimation, and labor bundling optimizer.

Launch Live App →
Opens in a new tab · Gemini 2.5 Pro + 2.0 Flash · Supabase · Zod

§ 03.3

Personal Project · Built in Bolt

Lumière Collective — AI Portrait Atelier.

A production-grade AI photo enhancement platform designed for a private luxury travel club's member directory. Upload or live-capture a portrait — receive four professionally styled variations in parallel.

Core concept

Members get four AI-generated portrait variations — Natural, Corporate, Editorial, and Vacation — in the time it used to take to get one. Four simultaneous Gemini API calls via Promise.all() reduce total wait time by ~75% versus sequential processing. The result feels like a concierge photo service, not a filter.

What's under the hood

Parallel AI Generation Google Gemini 2.5 Flash multimodal image generation, four simultaneous API calls via Promise.all() — each variation styled independently, results returned together.
Edge Function Proxy Bolt Database Edge Functions on Deno runtime handle API key management, server-side prompt safety validation with keyword/negation context parsing, automatic person-removal pre-processing, and CORS.
Live Camera & Upload React 18 + TypeScript SPA with live camera capture (MediaDevices API), drag-and-drop upload, interactive before/after slider comparison, and one-click download. HEIC/PNG/JPEG support.
Custom Luxury UI No UI library dependencies — hand-authored Tailwind CSS and CSS keyframes. Magnetic lagging cursor on requestAnimationFrame, kinetic typography animations, glassmorphism validation toast, Ken Burns hover effects, cinematic loading sequence.
Dual Safety Layer Client validation blocks disallowed requests immediately; a second server-side pass on the Edge Function prevents any circumvention.
Image Pipeline Client-side resize before upload, base64 encoding in chunked 8KB passes to avoid call stack overflow.
React 18TypeScriptTailwind CSSGemini APIDenoEdge FunctionsMediaDevices APIServerless
Live app →
luxury-profile-photo-ck11.bolt.host

Lumière Collective

AI portrait atelier — four professional variations generated in parallel via Gemini 2.5 Flash. Natural, Corporate, Editorial, and Vacation, delivered simultaneously.

Launch Live App →
Opens in a new tab · Gemini 2.5 Flash · React 18 + TypeScript · Deno Edge Functions

§ 03.3

Daily Practice

AI as a working environment,
not a feature.

This is harder to show than a shipped product, but it's probably more relevant to how I'd work on your team.

Claude + Cowork as a daily operating system

I use Claude Cowork — Anthropic's desktop AI platform — for a significant portion of my work: drafting PRDs, synthesizing research, building business cases, generating structured data, iterating on copy, and managing the cognitive overhead that used to require three browser tabs and a Notion doc. It's not experimental for me; it's how I work.

The practical result: I move from a fuzzy idea to a structured, reviewable artifact faster than most. That speed compounds across a team when you're the PM setting the pace.

ClaudeCoworkPrompt EngineeringLLM WorkflowsAI-Native

A note on this job search

I'm using AI throughout this process — for research, drafting, tracking applications, and building this site. Not because it's trendy, but because that's how I work now. If you're evaluating PMs on AI fluency, the most honest signal I can offer is that it shows up in how I operate day-to-day, not just in what I list under skills.