Skip to content

01  AI & Data Products  ·  Los Angeles

Models are easy.Adoption is thehard part.

I'm Richard — fifteen years turning data and models into products people actually open twice. Four enterprise GenAI systems in production, about 100,000 hours a year handed back to the humans who used to do that work.

Open to AI product leadership roles

Capital GroupAccenture @ FacebookSchlumbergerZKPnoteSupplyMind AI

02  Impact

What the work actually moved.

Measured in production, not in a deck.

0Enterprise GenAI products taken 0 → 1
0Hours a year returned to the business
0Data practitioners led at peak
0Years shipping data and AI products

03  Experience

The résumé, without the padding.

Product, platform and the teams that built both.

2026 — nowAI film studio

Founder & Product Lead — Hero Frame Media

  • Founded an AI film studio; a new YouTube channel reached about 590K views and 900 subscribers in its first five weeks, with single releases passing 100K on both YouTube and Instagram.
  • Designed a two-agent operating model, Claude and Grok, with written lanes, contract-style task requests and a human approval gate on every spend and every post; 47 delegated tasks run through it.
  • Directed the agents to rebuild heroframemedia.com as an interactive 3D studio (Next.js, Three.js, Postgres): concept to production in two days, across five reviewed releases.
  • Launched a follow-to-unlock content wall across four social platforms and a public live floor showing each agent's status.
  • Turned the lessons into reusable agent skills and a collaboration playbook.
2026 — nowGame, private beta

Creator & Game Director — The Last Archer Inu

  • Created and directed a browser action roguelite (React, Vite): six arenas, 10+ bosses with telegraphed attacks, four skill trees, 20 outfits and 60 bows, arrows and quivers, on desktop and phone. See archerinu.com.
  • Ran a three-agent build pipeline — planner, builder, deployer — with a written plan per feature, a staging build for playtests and my sign-off before every production release.
  • Designed the progression economy (tickets, star tiers, draw banners with pity); drop rates were verified against the design table over 200,000 simulated draws.
  • Shipped sign-in with X, a leaderboard API, cloud saves and a private-beta gate in front of the game.
2021 — 2026Los Angeles

Principal Product Manager, AI & Data Products — Capital Group ($3T AUM)

  • Launched Sales PrepMe, an LLM meeting-prep assistant over 20+ data sources — 15–20% less prep time, roughly 76,800 hours a year across 384 sales associates.
  • Shipped PCS Genie, the first GenAI assistant in Private Client Services: GPT-4 and RAG over 500+ internal documents, >90% response accuracy, ~20,000 hours saved a year.
  • Ran a four-product enterprise GenAI portfolio — RADAR, QAR Genie and Lighthouse — and wrote the value / feasibility / risk framework leadership now scores use cases against.
  • Delivered three "Golden Record" data products and the Model Alignment Index API; integrated 40+ third-party datasets and unlocked $29.3B of previously unattributed advisor AUM.
  • Cut project provisioning from 2–4 weeks to minutes on Databricks and Unity Catalog. Managed six PMs and analysts; founded the firm's continuous-improvement guild.
2017 — 2021San Francisco

Manager → Principal Consultant — Accenture @ Facebook

  • Head of Data Analytics on the Facebook account; grew the practice from a small engagement to 75+ analysts, engineers, data scientists and BI developers.
  • Led business development worth $20M+ a year across new and expanded client engagements.
  • Architected the governance and catalog program that turned data access into a shopping-cart checkout, with full ownership, lineage and workflow transparency.
2012 — 2017Houston

Data Engineer → Data Scientist — Schlumberger

  • Built pricing-anomaly detection end to end; the findings directly reset terms in client and vendor contract negotiations.
  • Shipped predictive maintenance on IoT logs for costly deep-sea solenoid failures, and surfaced $2M+ of annual overspend through license-asset clustering.
  • Migrated on-premise databases to AWS and Azure, designed a Cloudera Hadoop/HBase cluster, and built the data-quality framework that scored pipeline health.

B.S. Management Information Systems — University of Houston 1st place — Global Hackathon, 2016

04  Portfolio

Things I built on nights and weekends.

github.com/therealdealio — shipped, not demoed.

ZKPnote

01

Privacy-first notes with on-chain proof of authorship. End-to-end encrypted, SHA-256 content hashes anchored on Solana — about $0.21 a note against $5–15K for a traditional IP filing. Its 18-tool MCP server cut Claude token use by 98%.

Next.jsSolana / AnchorlibsodiumMCP

SupplyMind AI

02

An agentic iPaaS joining SAP to 400+ shipping carriers. Four Claude agents — Analyzer, Mapper, Transformer, Validator — and six parsers including EDI X12 and EDIFACT. A four-month manual integration becomes hours.

Next.jsClaudePostgresEDI

Learn Agent Architecture

03

An interactive LMS for agent design patterns — tool use, orchestration, evaluation — taught by the agents themselves. Full-stack TypeScript, built to be used rather than read.

TypeScriptNext.jsClaude

Hail Mary

04

A Polymarket trading agent running a nine-model pipeline — Bayesian updating, a Kalman filter, Kelly sizing, Monte Carlo — against live sportsbook intelligence, with a real-time dashboard over the top.

PythonFastAPINext.js

The AI PM Playbook

05

DARE and ALIGN: two frameworks for product managers walking into AI for the first time. Agent architecture, evaluation patterns and go-to-market, written for people who have to ship next quarter.

Open sourceFramework

The Prompt Codex

06

A prompt-engineering playbook for data organizations: nine roles, fifty-plus worked examples and twenty reusable templates, so a team stops reinventing the same prompt every sprint.

Open sourcePlaybook

Seedance Playbook

07

What five AI films taught us. Every prompt that worked, every one that didn't, and what each failure cost in render credits — the notes I wish existed before I started.

Open sourceAI video

bot-verify-web-feature

08

A Claude Code skill that writes and runs a browser test for every web feature it implements, fixes its own selectors until green, then hands the human a plain-English checklist.

TypeScriptPlaywrightClaude Code

Percolator

09

A permissionless perpetual-futures launcher: deploy a perp market for any Solana token in one click, with no listing committee in the way.

TypeScriptSolanaDeFi

05  Toolkit

Hands still on the keyboard.

A PM who can read the pull request.

AI & Agents

  • Claude, GPT-4
  • RAG & retrieval
  • Tool use, MCP servers
  • Evals & model scoring
  • Prompt engineering
  • AI governance

Product

  • 0 → 1 development
  • Strategy & roadmaps
  • OKRs and KPI design
  • User research
  • A/B testing
  • Go-to-market

Data & Engineering

  • Python, SQL, TypeScript
  • Spark, Databricks
  • Unity Catalog, Snowflake
  • dbt, Airflow, Kafka
  • AWS, Azure
  • Data modeling & governance

Web3 & Infra

  • Solana (Anchor / Rust)
  • Smart contracts
  • Cryptography (libsodium)
  • Next.js, Vercel
  • Postgres, Prisma
  • Docker, Git CI

06  Contact

Let's build something people open twice.