Beckett Dillon

Experience

1. Senior Forward Deployed AI Engineer and AI Consultant · Adobe

   Nov 2025 – now · Remote

   Embedded with enterprise customers and with Adobe’s own product teams: integrating Firefly Services, AEM and GenStudio into customers’ systems, and turning what the field taught us into internal tools, demos and product changes.

   • Baptist Health. Built Firefly panels for Illustrator (CEP) and Photoshop (UXP), released through the Creative Cloud marketplace as Adobe first-party plugins and in production with the design team, with custom models, vectorization and JPG and SVG pairs published to AEM; an agentic Premiere Pro panel that calls Firefly Services through MCP tools is in pilot
   • Baptist Health content supply chain. Helping build one intake-to-publish flow across Workfront, Creative Cloud and AEM with several pipelines attached; mapping it to the clinical service lines turned a 25-item conversion into about 250 repeatable deliverables, a 10x rescope
   • Delaware North. Wrote the pipeline spec and built a reference implementation on App Builder, a Frame.io custom action driving the InDesign API with an asynchronous webhook pattern for Frame.io’s 5-second limit, idempotent retries and server-to-server IMS auth. Their IT team built its MuleSoft pipeline against it
   • Logitech. Designed a two-path assets-at-scale architecture for a team with no in-house developers, a no-code workflow they own plus an API fallback across Firefly custom models, Substance 3D and GenStudio; found a Substance 3D compositing limitation and proposed a supported workaround
   • Fanatics Card Collectibles (Topps). Automating card production and design ideation at scale with the Photoshop API and Firefly Creative Production workflows
   • Field to product. Took customer feedback, API gaps, provisioning and permission blockers, and marketplace review requirements back to Adobe product and engineering teams as documented issues and proposals
   • Internal tools and demos. Built internal tools and enablement for the forward-deployed team, including a guide to provisioning LLM endpoints for prototypes; led customer demos, discovery workshops and prompt training for design teams

2. AI Solutions Engineer and Automation Developer · Doner

   Feb 2025 – Sep 2025 · Contract · Detroit, Michigan

   Turned brand, legal and production rules into pipelines for ACE Hardware, Stellantis, Tylenol and Rush Street Interactive campaigns.

   • ACE Hardware copywriter. A Dify pipeline that reads the team’s Excel and Word briefs, retrieves brand tone and style rules, checks SEO terms and returns structured copy for review, with Claude and GPT behind it. Copywriting time fell 60%
   • Stellantis. A four-phase After Effects pipeline in ExtendScript that builds slates, formats legal disclaimers with an LLM against brand rules, enforces safe areas on the timeline and flags off-spec spots, driven by CSV with a debugging UI
   • Rush Street Interactive. A multi-agent assistant that writes LinkedIn, Instagram and push copy per platform, using brand rules it extracts from the conversation; used in pitches and new business. Across accounts, campaign delivery ran 40% faster on reusable frameworks
   • Tylenol and signage. Generative video with Sora, Runway and ElevenLabs voiceover; Photoshop API signage templates filled from a copy bot, cutting manual input 70%

3. AI Engineer · Yamamoto

   Aug 2024 – Sep 2025 · Minneapolis, Minnesota

   Brought generative AI, AR and spatial computing into campaigns with the innovation team, alongside the Doner role.

   • Led prototyping and experiments with the Innovation Director, from concept to client demo, and presented in Adobe and Stagwell innovation sessions
   • Built AR and spatial experiences with generated content for campaign work
   • Scoped an LLM platform for Solventum

4. Founder, independent studio · Interwoven Arkitech

   Jan 2022 – now · Remote

   My independent practice for design, open-source AI, software and craft: tools I build to learn a new medium, and that other people use.

   • Open models. 46 models and 7 datasets on Hugging Face with 188,000 downloads, plus 138,000 on community re-quantizations by TheBloke, mradermacher and others; fine-tunes, mixture-of-experts merges and quantized builds
   • Open source. GraphRAG-Local-UI (2,321 GitHub stars, 291 forks) runs Microsoft’s GraphRAG on local models; Vodalus builds RAG-grounded fine-tuning datasets
   • ANIMA. A biomimicry agent and workbench where every number is computed or cited. It plans, runs tools in parallel, delegates to researcher, critic and engineer subagents, and audits its own citations over a 1,064-node knowledge graph, with 21 physics calculators and printable geometry
   • SlopSoup TV. A 24-hour channel run end to end by software, from writing and voices to music and the live YouTube broadcast, with critics at every stage, a daily spend cap, provider failover and a simulated 24-hour playout test
   • Model research. Ported Pathway’s Dragon Hatchling architecture from PyTorch to Apple MLX for about twice the training throughput; open notebooks on signal analysis and memory, drawn from how waves behave in natural systems
   • Apps, design and craft. Flint, an offline wilderness guide; Firewatch, a civil-rights companion in 13 languages; a training app; the visual system behind beckett-dillon.com; four-colour PETG prints

5. Advanced AI Quality Analyst and Data Trainer · Invisible Technologies

   Sep 2023 – Aug 2024 · Remote

   Trained and evaluated frontier language models with human feedback.

   • Reviewed LLM outputs, named the behavioral patterns behind failures and wrote natural-language feedback to steer them
   • Prepared training data for general-purpose and specialist models
   • Wrote quality benchmarks and evaluation rubrics used across training projects