SUNNYVALE, CA · GTM × AGENTIC AI

Marketing leader. AI builder. Same person.

I ship the AI I write about.

I'm Henrique Loyola. By day I lead Content & Discovery go-to-market for Google Play Games. Nights and weekends, I design, build, and run multi-agent AI systems on my own hardware — from e-commerce pipelines to Android apps. This site is the lab.

THE LAB

Things I've built and shipped.

Personal projects, built end to end on self-hosted models and my own infrastructure. Screenshots and demos below — these run in my house, not in a pitch deck.

Hg agent platform — multi-role session view

SHIPPED   STACK: Hermes · local LLMs · persistent memory · self-hosted

Hg

My agent platform: Hermes, running on local models on my own hardware, built around persistent memory. One agent, many roles — editorial drafting, research, turning photos of a dish into recipe cards, and holding my family's medical history for quick, private consultation. It remembers context across weeks, not chat sessions, and nothing leaves the house.

Levny — novel-writing studio dashboard

IN USE   STACK: Claude Code · local models · frontier editor

Levny

A novel-writing studio — chapter pipeline, canon and style guide, and a "review backwards" pass that walks a manuscript from the last chapter to the first, hunting continuity breaks. Built because the first book I needed to write was my grandmother's: she's 97, and I had hours of recorded interviews. Local models do the legwork; a frontier editor agent, fed samples of my own writing, takes the final pass so the draft reads like prose instead of a stitched transcript.

Merch pipeline — generated design gallery

SHIPPED   STACK: local LLM pipeline · print-on-demand

Merch Pipeline

An agentic pipeline that took a print-on-demand storefront from 5 designs to 500 — ideation, generation, listing copy, and upload prep — and reached revenue-positive within its first twelve weeks.

David Jobs — five-agent dashboard

SHIPPED   STACK: 5-agent system · Supabase

David Jobs

A five-agent career system that researches roles, tailors materials, and manages a search pipeline. Built for real use — including by people other than me.

JAM — application review screen

IN USE   STACK: Supabase · web app

JAM

A spin-off of David Jobs: paste a job description, and JAM reviews, scores, and tracks the application against it. Deployment-ready application management, minus the spreadsheet.

CivicAlignMap — neighborhood summary over map view

PRIVATE BETA   STACK: Google Maps · public civic data

CivicAlignMap

Built for my own family's next move: drop in an address or pan the map, and it consolidates the neighborhood picture — schools, local representatives, a one-line civic read from public records — into a single summary instead of fifteen browser tabs. AI consolidates the boring 80% of the research; the 20% that matters stays a conversation with real people. Friends and family only for now.

Looper — generated running route on map

PROTOTYPE   STACK: Android APK

Looper

A running-route generator for marathon training: give it a distance and a start point, it gives you a loop. Prototype APK running on-device, born from my own training plans.

Social network pipeline — draft review view

SHIPPED   STACK: multi-agent · human final cut

Social Network Pipeline

A multi-agent pipeline for social network content: one agent researches, one edits against my voice rules, and a frontier model takes the final pass. My posts about AI are produced by the AI they describe — and the last word is always mine.

EXPERIENCE

Two decades taking entertainment products to market.

My career runs through three platforms people love: movies, music, and games.

GOOGLE · 2017—NOW

Head of Content & Discovery GTM, Google Play Games. Led the zero-to-one launch of Google Play Games on PC and GTM for some of the largest mobile game launches in Play history. Today I lead a growing 12-person team and put multi-agent AI into real marketing production — including PEPA, Play's first culturally-aware multi-agent system.

APPLE · 2011—2017

iTunes Content Manager for Latin America, São Paulo — movies and TV across the region during the streaming transition.

EARLIER

Universal Pictures and Cinemark in Brazil: home entertainment marketing and film buying. I started this career choosing which movies theaters would show.

Full history on LinkedIn.

SPEAKING

Talking about AI in production, not in theory.

SHANGHAI · 2025

Google Think Games. Presented a study on AI-powered content localization for game marketing.

SILICON VALLEY · 2026

From Day One — Marketing Conference. Panelist on marketing leadership.

Henrique Loyola presenting the AI localization study on stage at Google Think Games Shanghai
THINK GAMES · SHANGHAI · 2025
Henrique Loyola speaking on a panel at the From Day One conference in Silicon Valley
FROM DAY ONE · SILICON VALLEY · 2026

For speaking inquiries, use the contact below.

ABOUT

Off the clock.

Brazilian-Italian-American — work authorization across the US, Brazil, and the EU — based in Sunnyvale, California, with my wife and daughter. Marathon runner working through the Abbott World Marathon Majors — Chicago is next. Lifelong gamer (Final Fantasy, since forever), Disney parks completionist, and the designated risotto guy at family dinners. Portuguese and English, natively; Spanish, enthusiastically.

Let's talk.

The fastest way to reach me: