Aaron Robinson

Aaron Robinson

Product, Strategy, & Analytics · Building with AI

Tech strategy and analytics professional. I excel in uniting cross-functional teams toward shared goals, deeply understanding user needs and product opportunities, crafting product strategies, and implementing roadmaps. Experienced in 0-1 products, consumer, and enterprise.

At Meta, I partner with CMOs and growth teams to elevate marketing performance, using advanced analytics and tailored product strategies. Previously I led consulting teams, pivoted a startup from restaurants to stadiums, and ran operations for a presidential campaign.

Love pickleball, travel (some favorites are the Galapagos Islands, the original Legoland in Denmark, Uruguay, and the Faroe Islands), and baking brown butter, triple chocolate chip, Nutella-stuffed cookies.

Past experience

Seven years at Meta in marketing science and measurement for the largest advertisers, plus strategy and analytics at Uber Eats, digital media products at Nielsen, public policy at Google, a stadium-payments startup (Parametric Dining), a co-founded ecommerce company (Bamboo Supply Co.), and state field operations on the 2012 Obama campaign.

AI Builder

How I think about AI

Two sides, always: building AI into the product, and building AI into how the company runs, horizontally across functions and vertically through the stack. I keep all of it anchored on a real outcome, like more revenue or time given back to the team, so computers do what they do best and people do what they do best.

The principles I run on:

  • Anchor to business value before novelty: Start where AI drives the most impact, a cost on the P&L or a real growth lever. Never build a feature because it's AI.
  • Let computers do what computers do best, and free people for what people do best: Sometimes AI replaces a task. More often it magnifies the person.
  • Taste and distribution decide it: The question isn't whether we can build it. It's what's worth building, and getting it actually adopted.
  • Maximize AI without putting the core at risk: Keep a hardened, reliable platform at the center and open a safe layer on top where the team can vibe-code and experiment. People build fast on the platform without ever being able to take the core down, rather than letting everyone build everywhere.

Areas I'm exploring and would pitch to companies to build or buy, to move past basic LLM chat:

  • A proactive assistant that surfaces the next move before you ask, not only when you open a chat.
  • An agent council that debates the work and hands you one clear recommendation.
  • A state-of-mind assistant that shifts work based on how someone is engaging.
  • Action-taking agents that go from advice to running whole multi-step tasks on their own.
  • Agent-ready interfaces that let an AI agent take actions directly, the same way a person would.

AI Projects

Encore - On This Day

Claude Code // SwiftUI // PhotoKit // Linear // TestFlight

A personal "on this day" for your own camera roll. It finds the photos you took on this day in past years, all on-device, with a vertical swipe and no ads. It's on TestFlight now, with an App Store release in review.

How it works: PhotoKit reads the camera roll entirely on-device, pulls the photos from this day in prior years, and presents them as a vertical flip-through with one-tap share cards. No servers, no accounts, no upload.

Built with: SwiftUI and PhotoKit, shipped through TestFlight. I skipped the usual PRD and described it to Claude Code, then refined it on my own iPhone through a loop: use it, file issues in Linear, Claude Code builds them, and TestFlight sends them back to my phone.

More information

Avails

Claude Code // Chrome Extension // JavaScript

Multi-timezone scheduling tool with AI-powered time parsing, natural language input, and 600+ IANA timezone support. Published on Chrome Web Store with side panel interface, smart conflict detection, and intelligent error handling.

How it works: Users type natural language like "3pm Tokyo, 9am London" and the extension parses input against 600+ IANA timezones, renders a unified view in a Chrome side panel, and flags scheduling conflicts automatically.

Built with: Vanilla JavaScript, Chrome Extensions Manifest V3, Chrome Side Panel API. Entire project vibe-coded with Claude Code from idea to Chrome Web Store publication.

Process: My standard path: brainstorm the idea with Claude Code, write a dual product requirements document (PRD) that is half human-readable and half technical spec, run an optional Claude AI design pass, then build and iterate.

Chrome Web Store GitHub

Summer Scheduler

Claude Code // Vercel // Supabase // TypeScript

Full-stack family scheduling app with conflict detection, gap tracking, and multi-plan management. Supabase backend with auth and real-time sync, deployed on Vercel.

How it works: Family members create and share scheduling plans. The app detects conflicts across overlapping plans, identifies unscheduled gaps, and syncs in real-time across devices via Supabase subscriptions.

Built with: Next.js (TypeScript), Supabase (auth, row-level security, real-time subscriptions), deployed on Vercel. Full-stack project built end-to-end with Claude Code.

Process: Same standard flow: a dual human-and-technical PRD up front, an optional Claude AI design pass, then a Claude Code build, with changes prioritized and shipped as I test.

Live App GitHub

Publications