2024 – Present
2–4 teammates
AI-powered code reviews for engineering teams
A shareware app to help you herd your MCPs
A community defining Ruby's place in AI
Sublayer is an agentic AI studio — a small team building software, frameworks, and communities at the intersection of AI and software engineering. I joined to help shape the product design across a growing portfolio of tools and community efforts, each one targeting a different slice of the developer workflow.
The challenge was consistent across all three projects: we were moving fast in a rapidly evolving space where the audience (developers) had high expectations for quality and low tolerance for fluff. That pushed every design decision toward being opinionated, clear, and purposeful.
AI-powered code reviews that enforce sophisticated, context-aware rules and stay effortlessly informed with smart weekly summaries that capture your team's progress across all projects, delivered exactly how you need them.
augmentations.ai ↗Augmentations started from a real pain point: code review is one of the highest-leverage activities in software development, but it's inconsistent and expensive. Reviewers miss things. Context gets lost. Standards drift. The product automates the enforcement of sophisticated, team-defined rules while also surfacing weekly progress summaries across all active projects.
My role was to translate a technically complex service into something approachable for engineering leads and developers who are already skeptical of AI tooling. That meant being direct about what the product does, avoiding marketing vagueness, and letting the depth of the feature set speak for itself.
The brand direction leaned into a deep purple palette — serious, focused, a little dramatic. Not trying to be playful. The layered box-shadow treatment on the launch card was an intentional nod to depth and weight; this is a tool that does real work.
Augmentations.ai marketing site
Project dashboard — Semantic Lint Rules and Generative Summaries
A shareware application bred to help you herd your MCPs. Manage, organize, and control your Model Context Protocol servers without losing the plot.
protocollie.io ↗Model Context Protocol (MCP) is how AI agents connect to external tools — databases, APIs, local files, third-party services. As MCP adoption exploded, the pain of managing multiple servers became real: which ones are running? What do they do? Where did that config go? Protocollie is a Mac app built to solve exactly that.
The naming and character design were central to the product's appeal. "Protocollie" is a Border Collie — bred to herd, naturally organized, relentlessly focused. The name is a pun that actually communicates the product's value, which is rare. My job was to carry that personality all the way through: the logo, the UI, the marketing site, the tone of all copy.
Shareware has a different design calculus than SaaS. You're not converting with a free trial — you're building enough affection and trust that people pay because they want to support the work. That means the product has to feel crafted, and the identity has to have genuine character.
Protocollie.io marketing site
A community of developers defining Ruby's place in AI. Bringing together Rubyists who are building with, thinking about, and pushing the boundaries of AI tooling.
artificialruby.ai ↗Ruby has a devoted community and a rich history in web development, but as AI tooling matured, most of the frameworks, libraries, and conversations were happening in Python. Artificial Ruby exists to change that — a community-first effort to keep Ruby relevant, adapt it to agentic workflows, and connect the developers already doing this work.
Community branding is different from product branding. The identity needs to feel inclusive and energizing rather than polished and exclusive. It needs to be something people want to put on their laptops. The "AR" mark leans into the duality of "Artificial Ruby" and "Augmented Reality" — a small nod to the space the community is exploring.
The challenge with community design is scope management: you're building something that lives across event materials, a website, social posts, and physical merch, without a dedicated marketing team. The design system had to be simple enough to be applied consistently by anyone on the team.
ArtificialRuby.ai community site
The meta-lesson from designing at Sublayer was about pace. In an agentic studio, the product roadmap can shift faster than a traditional design process allows. The tools themselves were evolving — MCP was barely standardized when Protocollie launched. Augmentations was being shaped in real-time by early-access user feedback. AI Ruby was growing through community pull, not a product plan.
That required a different design posture: be decisive early, invest in foundations (brand systems, component patterns, clear tone of voice), and then stay loose on the details. Shipping something good and iterating beats waiting for something perfect. The key is knowing which decisions are load-bearing and which ones aren't.
👋
