Complex software, made beautiful, built to scale.

Pathfinder is where technology advisors research, quote, and close enterprise deals, and it had grown faster than its design language. Three years of separate teams had layered five primary blues, three button shapes, and mixed type stacks into production, while marketing shipped at high volume with no shared foundations.

I didn't start with components. I started with an audit of every production surface. Once the mess was countable it was fundable, so the system had a business case before a single component existed. Advisors use this tool to make six-figure recommendations in front of clients. In that room, inconsistency reads as unreliability. That made it a trust problem more than a tidiness problem.

The system had to serve three things at once: a dark analytical product where advisors compare providers and close deals, a light high-volume marketing operation, and a live data layer of real revenue charts shared by both. Mapping those flows set the architecture early. Two themes resolving from one source, and components proven against the platform's densest screens rather than its simplest.

The spine that came out of it: one tracked project, five states. Advisors were running six-figure deals out of inboxes and spreadsheets. Mapping the work end to end gave the product a shape. Every screen answers for one state of one project, and every state is a different problem:

Project Information is where the advisor lives between deals. It has to answer "what needs me today?" in five seconds without becoming a wall of numbers. Configuration is the screen the client sees. It turns a market of dozens of providers into a defensible shortlist, one dense comparison row each.

Discovery is the longest state and the likeliest to stall, because assessments sit with engineers for days. The waiting had to read as clear progress, not anxiety. Registration is the irreversible step, which is exactly why the register control lives inside Configuration, next to the reasoning that justifies it. Contract is the state that has to outlive everyone. A signed deal becomes a deployment, then a billing relationship, then a renewal handled by someone who was never in the room.

Direction got decided in the open. Style tiles put real directions side by side, with palette, type, data density, and both themes on each one, so the team chose between options instead of arguing opinions. The tiles that lost still earned their keep. Every rejection sharpened the case for what won.

The model went to advisors before it went to build. The five states shipped first as a clickable prototype. The concept held. Advisors described the states back in their own words, unprompted, and stopped asking "where's the assessment I ran?", which was the question that had defined the old process.

The details didn't hold, which is the point of testing. The first Configuration table gave the fit score and the provider logo equal weight, so advisors scanned past the evidence and picked on brand recognition, the exact thing the screen existed to prevent. I rebuilt the row hierarchy until the score and the differentiator outweighed the logo. Registration moved for the same reason. I'd treated it as a light confirmation, advisors treated it as a commitment, so it went where the reasoning lives.

Every state, or it isn't done. Every component shipped with loading, empty, error, and disabled states. In a platform full of in-progress assessments, the in-between states are most of what people see.

The chart library carried the credibility: line, bar, area, stream, radar, and hex, all themeable, consistent, and legible with live revenue numbers on the screen in front of a client.

Three tiers of tokens: primitives, semantic, component. Engineers consume intent, never hex codes. Dark product and light marketing resolve from one semantic layer. Re-theme at the primitive layer and nothing downstream changes.

Governance built for adoption rather than enforcement. Contribution was federated. The docs were written for the person shipping at 4pm on a Friday. Weekly office hours caught edge cases before they became forks, and semantic releases came with migration notes so an upgrade never surprised anyone.

Systems prove themselves in shipped product. Components went to engineering as a React-ready library through Dev Mode, named, specced, and versioned so design intent survived the handoff. Clickable prototypes went in front of advisors before build, and when something didn't hold, it got rebuilt before it shipped.

Adoption was the metric that mattered. The library became the default path because it was the easier one, without a mandate.

Brand reviews shifted from policing basics to discussing ideas. New surfaces shipped visually consistent on day one. Accessibility findings stopped recurring because the fixes lived at the foundation level. And the architecture held: 2.0 to 3.0 on the same foundations, no rebuild. 3.0 is in development now, with a deeper dark theme and denser data visualization resolving from the same layer that shipped 2.0.

When the same feedback keeps arriving, the fix belongs upstream in the system, not in each deliverable.

After the product system unified Pathfinder, marketing was the next bottleneck: high asset volume, no shared foundations, and every campaign re-litigating the basics. The job was adapting the existing system for non-designers, who need harder rails and warmer documentation.

One template, a lot of room inside it. "Forge the Future" set the model: a locked typographic and brand structure with expressive visual fields, so campaign variants multiply without drift. Dozens of banner and social formats came from one template logic.

From campaign to calendar. The same template logic carried event creative and the social calendar. The NSP Advisory Board '24 package ran from stage screens to badges to email out of one Figma system, and holiday socials shipped on-brand without waiting on a designer.

Beyond banners. The library covered the unglamorous, high-volume work: 40-page industry reports, partner enablement, and event material, with templates carrying the brand's editorial layer so a report and a LinkedIn banner read as one company. The pattern throughout: I made the materials first (covers, reports, campaign assets), then refined each one into the library as a template the team could ship from.

Enablement over enforcement. Adoption ran through training sessions and self-serve guides written for non-designers. The training did more for scale than any template.

A worked example. The 2024 Colo & Cloud Report became the library's proof case for editorial work: a 40-page industry report designed on the same rails as the campaign and social kits.

Exploration inside the rails. Dozens of cover concepts in geometric, gradient, and fluid-wave directions, all drawn from the same brand palette, before the final was picked. The system constrained the exploration without flattening it.

Built to travel. The final cover shipped production-ready and co-branded for partner distribution, with interior templates carrying the brand's editorial voice across every page.

Refined back in. As with the campaign work, the finished report was distilled into library templates, so the next report started at eighty percent instead of zero.

The library became the default path because it was faster than the old way. Marketing stopped rebuilding basics, revision cycles on routine assets collapsed, and brand consistency stopped depending on any one person's vigilance, including mine.

A marketing library is a design system with a different user. Same architecture, different empathy.

BCBS member and enrollment experiences were built by many hands across many vendors, and it showed. Components were disjointed, specifications lived nowhere, and every new flow rebuilt the basics. For a healthcare brand whose whole promise is trust, the seams were the problem.

The build, end to end. Six moves from audit to adoption, and one rule underneath all of them.

Foundations first. A color system with interaction and accessibility tiers, a typographic scale built for dense insurance content, an icon library, and grids that worked from marketing pages to enrollment forms.

Accessibility as architecture. In healthcare, accessibility decides whether the product works at all. Contrast, focus, and plain language went into the foundation layer from the start instead of getting patched in after.

Proven where the stakes were highest. Plan selection and enrollment came first. Forms, payment, and confirmation patterns shipped as coded components with theming hooks, and the same foundations carried the marketing surfaces.

Documented like a product. Corner radii, modals, shadows, buttons and states, form fields, spacing steps, the multi-plan logo matrix, and responsive minimums, on one sheet a vendor could build from without a meeting.

The system in production. The under-65 shopping experience rebuilt as a conversation: one question per screen, plain language, and progress you can leave and come back to. Every screen below is assembled from FIBER foundations.

A system is only as good as its adoption. FIBER shipped with a guide I wrote and presented myself. It was built as an argument rather than a spec dump: here is the problem, here is the vocabulary, here is exactly how to build with it.

The brief, in the system's own words: "disjointed components with no real visible path to constructing the design to developers' specifications." The fix was a single source of elements and code that let any team produce mockups, prototypes, or production sites without error or question.

Shared vocabulary, communicated often. Designers, developers, and vendors named things the same way and pulled from the same sources. When a question came up, the guide answered it and the meeting never happened.

Beyond enrollment. The same foundations carried BCBS's public marketing surfaces, homepage through plan selection to payment, so the brand read as one product instead of a federation of vendors.

Teams stopped rebuilding the basics. Enrollment flows shipped visually consistent. The guide gave designers and developers one vocabulary, and concept-to-completion got faster across BCBS's product line.

Healthcare taught me accessibility discipline. I've carried that standard into every system since.

GE Healthcare's products span imaging, monitoring, and clinical workflow. This is software used in life-and-death settings, built by teams around the world. Edison unifies them with modular, accessible UI patterns that hold up in regulated environments and still flex from radiology workstations to touch-screen devices.

Edison was a team effort that predated and outlasted my engagement. My focus: scalable component libraries, theming, and documentation that let diverse product teams adopt the system without translation loss.

Four themes from one semantic layer. Clinical software runs in radiology reading rooms at 2am and on bright surgical floors. So: light and dark themes across desktop and touch targets, resolved from shared foundations so components never fork.

Components for regulated review. Notification patterns, steppers and progress, form controls, and the icon library, each with full state coverage and accessibility behavior documented for regulated review. When WCAG compliance is a legal requirement, accessibility has to be built into the architecture, because a checklist at the end won't pass.

Shipped as a kit. Edison went to teams as an installable system: GE typeface, Sketch libraries, theme files, and a written guide walking designers from setup through overriding symbols. The component menu alone reads like a product catalog.

Specified to the pixel. Every control shipped with full guidance on anatomy, states, spacing, and do-and-don't, written so clinical product teams could build without guessing. The checkbox guideline alone, lights-off theme:

The kit was half the job. Adoption was the product. Forty-plus teams around the world don't align because a system exists. They align when using it is easier than not using it. Edison's documentation hub became the single source of truth: one site for designers and developers, covering UI patterns, best practices, and framework-agnostic Web Components ready to drop into any product stack.

Libraries that keep themselves honest. Sketch libraries installed by theme and device tier (dark, global, light) kept every designer's local file in sync with the source. Updates flowed downstream without anyone chasing them, and version drift never got a foothold.

Documentation written for the person using it. Every page answered a working question: what is this, when do I use it, how do I override it. That's the difference between a system teams comply with and one they reach for.

The team's work earned the 2020 Red Dot Award for Interface Design, along with Fast Company Innovation by Design and the dmi:DVA. I played a central role in the system work behind that recognition, working with designers, engineers, and accessibility specialists across GE Healthcare.

The hardest problems on Edison weren't visual. They were getting dozens of teams to agree.

PRISM measures how people think, work, and relate. That data is complex and personally sensitive. The design problem: visualize it so a manager can act on it in thirty seconds without flattening the nuance or making anyone feel reduced to a chart.

Visualization as vocabulary. Relationship maps, trait spectrums, and team composites, each a reusable pattern with defined anatomy rather than a one-off chart. The same components scale from an individual profile to a team view.

Desktop depth, mobile moments. Desktop carries the analytical depth. Mobile carries onboarding, check-ins, and quick reads before a meeting. One modular kit and shared foundations, with the hierarchy re-tuned for each context.

Warm by design. Psychometrics get personal fast, so typography, color, and language did as much emotional work as the data itself.

PRISM Connect. The social layer: compare personalities, processing styles, motivation, and conflict management across your network, on any device.

Chart anatomy, treated like component anatomy. Every visualization got the same spec a button would: defined states, sizes, and usage rules. That's the difference between a dashboard that reads as one language and one that reads as chart soup.

Approachable by design. The Portrait onboards users one step at a time, a guided reveal from Primary Personality through Decision-Making and Conflict Management, each with plain-language strengths and a short video, so dense psychometric data never reads like a test.

Relationship Advisor: data becomes guidance. The system's payoff. Side-by-side psychometrics translated into concrete advice for working with a specific colleague, so instead of "here's your score" you get "here's what to do in your next conversation."

Navigable, not overwhelming. Anatomy, states, and hierarchy designed as a system are what keep dense behavioral data legible at a glance. That's the thirty-second read the platform promised.

One UI framework across the full platform, with interaction models and information architecture defined for complex data visualization, and modular kits the team could keep building on after I left.

Data visualization is a component problem. Define the grammar once and every chart after that is easy.

TUCM started the way every honest system does: with an audit. I mapped every component, color, and pattern in production into one research document that made the fragmentation undeniable and gave the system its priorities. The audit doubled as the business case. Once the fragmentation was countable, the system was fundable.

The system. Fully responsive symbol libraries (this was years before Figma variants, so it meant disciplined Sketch architecture) with foundations tuned for data-dense credit dashboards.

Documented as a product: the SCVD guide. The Standard Credit View Dashboard shipped with its own written system guide covering libraries, color, typography, icons, buttons and inputs, and plugins, built so a partner team could take it and run.

White-label by design. Change the generic logo once on the symbols page and it updates on every screen in the document. Colors and type re-theme the same way. One system, any partner brand.

A system is only proven when it ships a product. TUCM's proof was TrueCredit, a full rebrand and design pass on TransUnion's consumer credit experience, built entirely on system foundations. Credit scores, calculators, and monitoring walkthroughs were assembled from the same symbols and libraries the guide documented.

One system, every screen size. Because the symbols were responsive from day one, the TrueCredit family (dashboard, calculator, pricing) moved from desktop to mobile without a redesign pass. The breakpoints were already in the components.

People have to believe the number. Hierarchy and plain numbers decide whether someone trusts their own credit score. Building the consumer product on the system made that discipline structural instead of a per-screen style choice.

Design-to-code workflows got a shared vocabulary. The marketing library unified the brand touchpoints. Credit data, some of the most stressful content on the internet, got an interface language built for clarity. The pattern of audit, then argument, then architecture has carried into every system since.

People flinch at credit data. Clear hierarchy and plain numbers are what let them believe it.

Credit data is some of the most stressful content on the internet. TrueCredit needed a brand and product language that made a score feel like information instead of judgment: friendly enough for a first-time user, precise enough to be trusted with $149,316 of credit balances.

Five visual directions pushed the rebrand from safe to decisive: photography under brand marks, isometric illustration, saturated color fields. Each was carried across desktop and mobile before one was chosen.

The chosen direction carried into the full product: walkthroughs and a credit calculator for first-time users, credit lock, reports, and theft alerts for daily monitoring, and a free-to-premium pricing path. Every screen was paired desktop and mobile from day one.

A consumer credit product with a face people didn’t flinch at. Built on TUCM foundations, so the rebrand proved the system while the system carried the rebrand.

People flinch at credit data. Clear hierarchy and plain numbers are what let them believe it.

Maven Wave needed a visual brand that could hold a hard middle ground: professional, high-quality, innovative, likable, and smart. A traditional consulting firm on one side, a cloud brokerage and technology firm on the other. The identity had to show real capability and leave room to grow.

The presentation opened by teaching the client how to judge a logo: Paul Rand, the Citi case study, and the rule that a mark earns meaning by association. Simple, memorable, timeless, versatile, appropriate.

Each concept was built out as a complete system rather than a sketch: construction logic, mark, lockups on light and dark, a typeface cast to match the personality, and a physical mockup to make it real.

The ribbon wave: M and W woven into a rising pulse, set in Rotis Semi Serif. The flux sphere: motion and orbit, with its own modern grotesque. The chevron cube: an expert’s M built from dimensional planes, set in Cosmos. The loops: interconnection between consultants and clients, set in Barmeno.

Four directions, presented as one argument about what a brand is for. The lineup gave Maven Wave a real decision to make, with mark, voice, and material presence considered together. The firm went on to become one of Google Cloud’s premier partners and was later acquired by Atos.

A logo gets its meaning from the quality of what it stands for. The deck taught that principle, then proved it four ways.

A logo gets its meaning from what it stands for. The job is a mark that is simple, memorable, timeless, versatile, and appropriate, and then getting out of its way. These twenty are shown as delivered: no mockup theatre, just the marks.

Before a mark, a name. The exploration ran wide on purpose: Cozzi, Adore, Halos, Lake House, Campfire, Clean Harvest, Tastelab, Varsity and a dozen more, each drawn as a real lockup and staged in context, because a name can only be judged wearing its typography.

The full deck, page by page. Thirty-eight boards, exactly as presented. Every name staged in context before the field narrowed to one.

Spacelab won because it carried the promise: science meets delicious, precise dosage with a playful spirit. The astronaut-helmet mark and stencil-cut wordmark held up from a foil stamp to a storefront, and the system flexed from lab-clean white to deep-space color.

The mark had to survive every format the line would ship in. One celestial artwork language (star fields, orbital arcs, the helmet) carried across drawer boxes, square boxes, tins, and jars, with the wordmark and its clean-label line holding the same position on every face.

One structure, many colorways. The same drawer box, re-skinned: midnight blue for the core line, burnt orange for the seasonal. Structure, type, and lockup never move. Only the field behind them changes.

Down to the small formats. Tins and jars get the same treatment at a fraction of the surface, which is the test any identity has to pass.

And a second artwork direction. Running alongside the star fields: flat geometry, with half-circles, wedges, and triangles in ochre, slate, and red. Same helmet, same black label bar, same rules. It became the face of the line.

A complete shelf presence from nothing: name, mark, system, and packaging line speaking one language. The exploration deck did its job twice. It sold the direction, and it documented every road not taken.

The decided system, one board. Geometric pattern in slate blue, ochre, and burnt red, with the helmet mark and black label bar holding every format (tin, jar, pouch, tube, and box), down to the helmet embossed on the gummies themselves.

A name can only be judged wearing its typography, so every candidate got dressed before the vote.

The brief was self-set: what would a Google-grade information dashboard look like if density were treated as a design material instead of a problem? The design works in cards and charts with strict hierarchy: green-to-blue status language, generous counters, and data visualization that reads at a glance on desktop and tablet.

Nobody asked for this one. It still shapes how I draw charts today.

In 2010, putting a full client-relationship tool in an agent’s pocket was the frontier. The design paired a desktop workspace with a companion mobile app: policy lookups, claim status, and client management in a visual language matched to one of America’s most recognizable brands.

Early proof of a habit that stuck: understand the workflow first, then pick the device.

Porter treated hair care as a luxury service, and the app had to feel like the lobby: warm photography, disciplined type, and a booking flow with zero friction. The interface put stylists and products a tap away while keeping the brand’s premium hush intact.

The screens stay quiet on purpose. The service is the luxury, and the interface just holds the door.

Years before responsive frameworks made it routine, this storefront put Dremel’s full tool and accessory catalog on the phone. You could browse by project as much as by product, because makers start with the thing they want to build.

Letting people shop by project instead of spec turned out to matter well beyond tools.

Identity and interface shipped as one system: a sharpened mark, a site rebuilt around it, and product screens that inherited the same geometry and palette, so the brand didn’t stop at the marketing page.

Plenty of rebrands stop at the marketing page. This one carried through to the login screen.

After the 2017 breaches, freezing your credit went from an obscure safeguard to something everyone was doing. This ideation explored how a freeze product could look approachable instead of bureaucratic: friendly geometry, clear states, and a visual temperature that says protected rather than punished.

Security products need to feel calm. The ideation aimed the whole brand at reassurance.

Arrow’s brand is industrial confidence, and the storefront leaned into it: big product photography, unmissable red, and a catalog structure that gets a contractor from need to checkout in as few taps as the hardware aisle has steps.

Contractors don’t browse. The design gets them from need to checkout and back to work.

A startup selling a new idea needs its story drawn as much as written. The rebrand paired a fresh mark and site with a system of infographics that explained the skills network visually, charts and diagrams doing the elevator pitch’s heavy lifting.

When the product is a new idea, the infographic is the demo.

Financial services live or die on perceived competence. The rebrand gave Devine a composed mark, a site with unhurried typography, and comparison infographics that made complex offerings scannable.

In finance, restraint is the brand, so the design spent its energy on being believed.

Pharmacy benefits are where healthcare complexity hits the wallet. My Catamaran translated it: refills, coverage, and cost transparency in an app with a friendly identity and screens a member could parse in a waiting room.

An early rehearsal for the healthcare systems that followed: plain language as a design requirement.

Before the smart home settled into platforms, Vision imagined the appliance as a teammate: cycle status, efficiency scores, and camera views in a companion app with Whirlpool’s trusted warmth rather than gadget cool.

The appliance already had the data. The app’s job was making it feel domestic.

The 1,000 Greetings anthology collects standout correspondence design from studios worldwide. This holiday suite of navy, ornament pattern, and metallic accents made the cut, credited under Firebelly Design with design by Mikel Rosenthal.

Print is the one medium with a spine. This work sits on shelves, not servers.