I treated my job search like a product problem

At first, I thought I needed better bullets.

What I really needed was a clearer professional story.

My old resume had plenty of experience, but it asked the reader to connect too many dots. B2B SaaS sat next to ecommerce. Product design sat next to digital marketing. Founder work overlapped with full-time employment. Some of my most ambitious Sakari projects never reached production, while Course Vaults had shipped but didn't yet have a conventional job title attached to it.

None of those things were necessarily weaknesses. Together, though, they created ambiguity.

Before rewriting anything, I worked with AI to build an inventory of what I had actually done. We went role by role, separating responsibilities from outcomes and shipped work from exploratory work.

That meant digging through details I hadn't thought about in years:

• Sakari's Arctic design system had grown to nearly 1,000 components and variants and was adopted across the product.
• I had worked closely with two front-end engineers through Figma, Storybook, and Chromatic.
• Pheroe generated approximately $300,000 over three years before an asset sale.
• At Omura, I designed the Shopify experience, navigated payment-processing restrictions, built subscription and lifecycle programs, and helped grow online revenue from zero to approximately $15,000 per month.
• At Murchison-Hume, I worked across acquisition, conversion, ecommerce, and subscriptions.
• Course Vaults had grown to more than 7,000 registered profiles, nearly 40,000 courses, more than 226,000 course ratings, paid subscriptions, and meaningful organic search visibility.

The exercise surfaced something important: my experience wasn't unfocused. It had a consistent thread.

I tend to work in small teams, take ownership of loosely defined problems, connect product decisions to commercial realities, and stay involved until something becomes real.

The next question was whether one resume could tell that entire story.

Technically, yes. Strategically, probably not.

A hiring manager looking for a Staff Product Designer cares about different evidence than a founder hiring their first designer. A growth team will interpret the same experience differently again.

We landed on three primary career lanes:

This version emphasizes complex SaaS, systems thinking, product architecture, design systems, cross-functional leadership, and mentorship.

Sakari provides the strongest evidence here, particularly Arctic's adoption across the product and my experience operating as the senior and primarily sole designer on a small product team.

This version emphasizes zero-to-one ownership, ambiguity, product strategy, rapid prototyping, monetization, and the ability to move between customer needs, business constraints, and implementation.

Course Vaults leads this story because I founded, designed, launched, monetized, and continue to operate it.

This version emphasizes acquisition, onboarding, activation, conversion, subscriptions, lifecycle marketing, ecommerce, and revenue.

Course Vaults, Pheroe, Omura, and Murchison-Hume all contribute evidence to this version.

Behind those three sits a master resume: the complete source of truth for my experience. The targeted resumes are lighter layers on top of it, not completely separate identities.

Roughly 80% of the information stays consistent. The headline, profile, capabilities, emphasis, and selected bullets change depending on the role.

That distinction matters. The goal isn't to manufacture a different candidate for every application. It's to help the reader see the most relevant version of the same person.

At one point, I considered creating a separate AI-native resume.

I use Cursor to build Course Vaults across React, Expo, Supabase, and Vercel. I've designed multi-agent workflows to help clean and enrich course data. AI has significantly changed how I prototype, validate ideas, and move from a product decision to working software.

It seemed different enough to deserve its own category.

But when I started reviewing actual job listings, that model broke down. AI-native expectations appeared across Staff, Founding, and Growth roles. AI wasn't replacing those disciplines; it was becoming part of how the work gets done.

So I changed the system.

The three career lanes stayed intact, while AI became a second dimension:

• Standard: AI isn't central to the role.
• AI-forward: Modern AI fluency is expected, but the job remains primarily product design.
• AI-specialist: Prototyping in code, agentic workflows, or AI product experiences are central to the position.

That also changed the resumes. Instead of isolating AI in one experimental version, I wove it into all three, using different evidence and emphasis depending on the role.

It was a small taxonomy decision, but it made the entire system more coherent.

Once the positioning was clearer, the next problem became discovery.

Job boards make it easy to find a lot of roles and surprisingly difficult to maintain a high-quality search. Titles are inconsistent. Posting dates are unreliable. Remote policies are vague. The same role appears in multiple places. And after enough scrolling, almost anything can start to look relevant.

I wanted a system that could continuously find roles while respecting a fairly strict definition of fit.

So I created a small agent-assisted operation.

I interact with a chief-of-staff agent called Cora, which helps manage the larger process and keeps track of the rules. A specialist agent called Scout searches for roles, verifies the original listing, and adds legitimate matches to a shared Notion queue.

The important part isn't that the agents have names. It's that they have clearly separated responsibilities.

• Scout discovers and evaluates opportunities. It does not apply.
• Cora manages the criteria, surfaces questions, and helps me review the queue.
• I remain responsible for deciding where to apply, tailoring the story, and submitting the application.

That separation prevents a common automation problem: allowing the system to take action before its judgment is trustworthy.

Before automating anything, I had to make my own preferences explicit.

A role must map to one primary career lane. It also receives an AI-relevance tag. Listings must be recent, connected to a real company source, and compatible with my location.

I'm based in Napa, so remote roles are preferred and Bay Area hybrid roles are workable. Full-time onsite roles and hybrid positions outside the Bay Area are excluded.

The queue only contains roles worth considering. Rejected jobs aren't added simply to demonstrate that the system found something.

That sounds obvious, but it changed the quality of the workflow. The goal isn't to produce a busy dashboard. The goal is to reduce the number of bad decisions I need to review.

The system also maps each role to evidence from my actual background. It can't call something a strong fit without explaining why.

That requirement helps prevent the language from drifting into generic AI confidence, or overstating work that never shipped.

When I decide to apply, the job is matched to one of the three base resumes.

The application version then receives a light adjustment:

1. Match the headline to the role's actual title.
2. Update the three positioning tags beneath it.
3. Refine the profile and capabilities where necessary.
4. Reorder or adjust a small number of bullets.
5. Export the application copy as a PDF.

For example, a Lead Product Designer, Enterprise role starts from the Lead/Staff resume and emphasizes complex workflows, design systems, leadership, and B2B SaaS.

A Senior Product Designer, AI-first Support role can use the same base resume but emphasize AI-assisted development, messaging workflows, rapid prototyping, and product and engineering collaboration.

The underlying experience doesn't change. The entry point does.

I took the same approach with cover letters. The strongest version is still written in my voice and led by the problem I can help solve. The AI system is evidence of how I work, not a gimmick that speaks on my behalf.

The system now includes:

• A complete master inventory of my work and supporting evidence
• Three targeted resume lanes: Lead/Staff, Founding, and Growth
• AI positioning integrated across those lanes
• A shared job-prospect queue in Notion
• A specialist agent that searches for and evaluates current roles
• Clear filters for freshness, location, fit, and duplication
• A lightweight process for tailoring resumes and cover letters
• A human review step before any application is submitted

The discovery and review workflow is live. Fully automated application submission is intentionally not.

I want the system to become trustworthy before giving it more authority.

The biggest lesson wasn't about agents or automation. It was that AI becomes much more useful when the human has done the work of defining the system.

The quality came from deciding:

• What kinds of roles I genuinely want
• Which evidence supports each one
• Where my experience is strong
• Where it is ambiguous
• What the agents are allowed to do
• What still requires human judgment

AI helped me examine my background, find patterns I had undersold, challenge weak positioning, create multiple representations of the same evidence, and turn those decisions into a repeatable workflow.

But it didn't invent the strategy for me. The strategy emerged through iteration: questioning the categories, checking them against real jobs, and changing the system when the model stopped making sense.

That's also how I like to approach product work.

Start with the messy reality. Define the objects and constraints. Build the smallest useful system. Keep judgment in the right places. Then make it faster.

I started with a resume I wasn't confident represented me and no clear answer to what I should apply for.

I ended with a much clearer view of my professional value, three credible career directions, a reusable resume architecture, and an application workflow that can keep working without turning job hunting into my full-time job.

The system is still evolving, and the real outcome will be the quality of the conversations and opportunities it creates.

But even before that happens, the process has already done something valuable: it helped me explain my career in a way that finally feels accurate.