AI Systems Engineering vs AI Email Design: Which Skill?

What does Burtenshaw's AI Systems Engineering via Coding Agents do?

This skill pushes coding agents beyond routine software tasks into genuine AI and ML systems engineering. It is organized around three progressive tiers of difficulty, called "Bosses." Boss 1 is writing and optimizing custom CUDA kernels for specific GPU hardware — work that targets memory bandwidth bottlenecks rather than raw compute. Boss 2 is zero-shot LLM fine-tuning, where a plain-language instruction triggers the entire training pipeline on Hugging Face Hub. Boss 3 is AutoLab, a fully autonomous multi-agent research system with four specialized roles — Researcher, Planner, Workers, and Reporter — that propose, run, and evaluate ML experiments in parallel.

The core philosophy is "go closer to the silicon." Routine coding is commoditized; kernel optimization, training automation, and distributed research loops are not. The skill relies heavily on file-based "Skills" (structured context documents with examples and benchmarking scripts) that convert agent tasks from zero-shot to few-shot, dramatically improving output quality. Everything runs on open primitives — Parquet files, Git repos, CLI tools — because any abstraction an agent cannot inspect is a hard ceiling.

This skill is clearly for ML engineers, AI infrastructure teams, and researchers who need to squeeze performance out of hardware, automate model training, or run systematic experiments at scale.

What does the AI Email Design System do?

This skill enables anyone — marketer, brand operator, or solo founder — to produce a complete, editable, high-converting email design in under 10 minutes without a design team. It uses Claude's Design System or Design Project features as the primary tool, supplemented by ChatGPT for hero image generation when needed.

The workflow starts with gathering brand assets (website screenshots, logos via Brand Fetch, product images with backgrounds removed) and 3–4 reference emails from Milled.com. You then create a reusable Design System in Claude, upload everything, and submit a brief that explicitly includes your high-converting email formula — the structural sequence of hero visual, headline, ingredient highlight, benefits section, and CTA. Claude asks clarifying questions, generates the email, and lets you directly edit sections without reprompting.

The key insight is the "mix-and-match platform strategy": ChatGPT generates better hero visuals faster, while Claude produces better full editable email structures. Combining them yields the best result. The skill positions AI as a foundation that removes execution bottlenecks, but emphasizes that strategic input — knowing which formula to apply, which audience to target — remains a human responsibility.

How do they compare?

These skills operate in entirely different domains with zero functional overlap. Burtenshaw's skill is infrastructure-level AI engineering: GPU kernels, model training, autonomous research pipelines. The AI Email Design System is a creative production workflow: brand-consistent email marketing assets.

The complexity gap is significant. AI Systems Engineering requires deep ML knowledge, GPU hardware access, and comfort with coding agents, multi-agent orchestration, and Hugging Face's ecosystem. The Email Design System requires marketing judgment and basic familiarity with Claude and ChatGPT — no coding, no ML expertise, no specialized hardware.

Time-to-value also differs dramatically. A complete email design takes under 10 minutes. A CUDA kernel optimization cycle takes hours; an AutoLab research run may execute overnight. The outputs are fundamentally different: one produces quantitatively benchmarked ML artifacts (kernels, models, experiment rankings), the other produces visually polished, table-based HTML emails ready for deployment.

The one conceptual thread they share is the use of structured context to improve AI output quality. Burtenshaw uses file-based Skills to move agents from zero-shot to few-shot. The Email Design System uses Design Systems loaded with brand assets and a documented conversion formula to move Claude from generic to brand-specific. Both argue that the quality of your input system determines the quality of AI output.

Which should you choose?

Choose Burtenshaw's AI Systems Engineering if you are an ML engineer, AI infrastructure engineer, or researcher. Specifically:

- You need to optimize inference cost by writing custom CUDA kernels for specific GPU hardware.

- You want to automate LLM fine-tuning without manually writing training scripts.

- You want to run systematic, parallel ML experiments autonomously using a multi-agent research team.

- Your work involves verifiable, quantitative outcomes (speedup percentages, validation loss, bits-per-byte).

Choose the AI Email Design System if you are a marketer, e-commerce operator, agency professional, or anyone who needs email designs:

- You need a promotional, product launch, or subscribe-and-save email design quickly.

- You lack an in-house design team or want to accelerate ideation before handing off to designers.

- You want a reusable brand engine that produces consistent designs across sessions.

- Your work involves visual, qualitative outputs reviewed against a conversion formula.

There is no use case where these two skills compete. Your job function and the problem in front of you make the choice obvious. If you are writing CUDA kernels, the Email Design System is irrelevant. If you are launching a product email campaign, AutoLab is irrelevant. Pick the one that matches your work.