Bill Moore

GENE — Conversational AI at BCG

Bill designed the prompt architecture, personality, and voice for BCG's conversational AI agent — then produced and co-hosted 52 podcast episodes alongside it, referred to on-air as "BCG's AI Whisperer." Built on a 3-person team starting July 2023. Bill created the three-layer prompt system (context/conversation/suffix) that kept GENE coherent across two years of production, directed the engineering of the GENE|OM platform, and eventually took over development. GENE conducts live interviews at events including Davos, Cannes Lions (2024), and WIRED/Condé Nast. First non-human entity to win a Webby Award for Best Podcast Cohost (2024). Featured in Forbes (twice), BBC, and The Sunday Times.

Mural — AI Product Vision

Worked directly with Mural's CEO to define the vision for the future state of their SaaS collaboration platform. The problem: how does a visual collaboration tool evolve when AI agents can participate in the work, not just assist with it? Built a working prototype of a future-version agentic UI and produced a vision video to align the executive team around a product direction. Rapid prototyping from concept to clickable demo to cinematic presentation in eight days.

Deck-to-Video Production Pipeline

Bill designed and built a browser-based workflow automation tool that converts client presentation decks into rough video cuts through an eight-step automated pipeline: deck ingestion, creative brief extraction, script generation, storyboard generation, voiceover synthesis, timed animatic assembly, human review, and loop-back refinement. The human shows up at step 7 — the system handles the other six. From deck to rough cut in minutes. Used as the live demo centerpiece for Bill's AI Systems Design workshops.

Narrative AI Platform — Demo & Production Pipeline

Bill designed and built the interactive demo and full production pipeline for a venture-backed narrative AI platform. Sole creative technologist on the engagement. The demo is a voice-driven interactive fiction experience he built end-to-end: AI narrator with dynamic state control, voice conversations with AI characters, and a branching plotgraph engine. He also built the production pipeline that generates investor video assets — AI-generated visuals, voice synthesis, config-driven audio mixing, and automated editing handoff to Premiere Pro.

SUPERWORLD!

Bill designed and built an AI narrative game with a hidden director system. Claude acts as omniscient narrator who becomes any character the story needs. No mode switching, no player triggers — the narrative decides. Bill created every layer: game design, prompt architecture, visual design system (1-bit Macintosh aesthetic, ChicagoFLF font, Atkinson dithering), a custom nanoGPT model (10.85M parameters trained on 5M characters of game corpus), Flask backend, AI image generation pipeline, and ElevenLabs TTS integration.

Independent Consulting · 2012–present

Bill builds AI systems and enables teams to build their own. Workshops are how many client engagements begin — he teaches teams how the technology works, they build working software during the session, and they discover both what's possible and what they need built next.

Hands-on curriculum for designers, strategists, and business operations teams — people who have never opened a terminal. Six modules covering APIs, SDKs, open source repos, Git, spec-driven development, agentic CLI tools (Claude Code, Cursor), systems architecture, and security and hosting. Participants leave with a working application and a mental model for how AI systems are designed.

Delivered to Portrait Media (co-led with Mickey McManus), BCG design teams, and client organizations across automotive, consumer goods, and government.

Regional Creative Director, BCG Design Studios

Eight years designing AI applications and media experiences for major global brands. Works in a small team building client-facing generative AI tools, interactive prototypes, and production pipelines. Operates as a hands-on creative technologist — designing the system, then building it.

Founded BCG's generative AI community in April 2022. Started as a 10-person Zoom workshop, grew into #creative-technology-club: 1,354 members across BCG Platinion, BCG Global, and TDA. Published internal guidance on AI-generated content for client deliverables.

Broadcast Producer, Brunnerworks

Directed and produced broadcast and digital video content across agency clients. Directed the award-winning Huffy 125th anniversary brand film and television spots (Silver ADDY Award, AAF Pittsburgh 2018) — the anniversary video convinced Huffy's incoming CMO to take the role. Supporting production team for 84 Lumber's Super Bowl LI commercial (2017), which generated 6 million web requests in its first hour.

Media Faculty, The Art Institutes

Taught cinematography, editing, and production courses in the Digital Media program.

Director and Senior Editor, Wrecking Crew Media

Directed and edited Spotlight on Design, a multi-episode video series for SAE International covering automotive engineering. Sold as subscription content. Directed and edited award-winning television spots and video content for Pittsburgh Penguins, Heinz, Pittsburgh Pirates, Pittsburgh Technology Council, and other regional and national brands. Five Telly Awards (2013–2014).

First, Do No Harm

Bill's original research: when prompt engineering helps, hurts, and doesn't matter. Multi-model benchmark study across Claude, GPT-4o, GPT-5.2, Gemini, and Haiku. Experiment code, data, and full paper published on GitHub. Bill built Bench, a custom research instrument with three AI reviewers (Skeptic, Methodologist, Practitioner) who disagree with each other.

AI Coding Rules

12 rules that prevent AI coding disasters. Derived from 500+ hours of real production work. Works with Cursor, Claude Code, Windsurf, and GitHub Copilot. Includes auto-install script and rule files for every major AI coding tool.

Neural Art & Early Diffusion (2020–2022)

Photographs processed through neural networks and early diffusion models beginning in 2020. Work with neural style transfer, Disco Diffusion, and early generative tools. Named after the Richard Brautigan poem.

The Sorkinizer (Late 2019)

A GPT model fine-tuned on West Wing scripts that generates complete episodes in Aaron Sorkin's voice. Full episode — "We the People" by "Aiaron Sorkin" — published as a demo. Early exploration in style-specific language model training.

The Wave Distorter — High-Temperature Language Research

Bill's three-year research program exploring high-temperature language model output as a creative medium. "The least likely tokens are the most interesting ones because they're the ones the model was trained to suppress." He built a three-stage pipeline: base language models (GPT-2, nanoGPT) at temperature 1.5+ generate raw signal, a glitch formatter introduces Unicode corruption and multilingual injection, and Suno Bark TTS transmits the result as audio. Bill built a physical prototype from a vintage telephone handset connected to a Raspberry Pi. Demonstrated live at WVU. Produced an invented lexicon of words pulled from latent space.

Generative Video Studies (2022)

A series of studies combining generative audio and video models. OpenAI Jukebox generating music paired with BigGAN visuals. VQGAN "machine fever dreams." Stable Diffusion seed travel animation — a technique Bill developed independently for creating motion from incremental seed changes (2022).

STEF — Persistent AI Systems Research

Bill's ongoing research in making an LLM-based agent that persists across sessions, maintains identity, and operates autonomously. He built the full system: personality engineering that survives context window compaction, autonomous tool orchestration (Gmail API, iMessage, calendar, GitHub), scheduled operation via launchd, and memory architecture that accumulates knowledge across hundreds of sessions. 25+ shell scripts, Python tools, and monitoring infrastructure — all designed and built by Bill.