AstrumU

Overview

AstrumU was building an AI data platform to power a skills-based economy — ingesting transcripts, resumes, and work histories; translating them into verified skills profiles; connecting those profiles to workforce data. The technology was real. The knowledge graph worked. The company had a translation problem engineering couldn't solve.

Leadership conversations circled. Teams carried different interpretations of what the platform was for. An internally popular skills heatmap showed aggregate density across populations but couldn't answer so what? — interesting data isn't actionable insight. The work was collapsing the abstraction: make value legible for one person, then use that clarity to focus the roadmap.

Role

Head of Design, operating in a hybrid design–product capacity — Seattle, WA · 2024–2025

• Owned design vision and shared product strategy with senior leadership
• Built and managed a 4-person design team across Product and Marketing
• Led AU Labs from concept to the company's primary alignment, investor, and customer-positioning tool
• Founded AstrumU's first qualitative research practice, embedding discovery into every sprint
• Defined the product catalog, OKR/KPI frameworks, and Product Readiness Tracker adopted across departments
• Partnered with Product leadership on the career-mobility wedge decision

Approach

Platform Direction

Built AU Labs as a narrative arc that answered "what does this platform do for one person?" — then used the clarity to pick the wedge that needed no translation. AU Labs was a six-step experience walking a talent-management stakeholder through the platform's value, comparison-driven rather than claim-driven. It proved the knowledge graph by showing outputs side-by-side with alternatives, not by asserting superiority. With the unified profile anchoring the narrative, four candidate use cases — education ROI, hiring efficiency, workforce planning, career mobility — narrowed to the one where the profile was the product.

Impact: The executive conversation shifted from "what are we building?" to "how do we get this in front of users?" Investor decks and customer-positioning materials inherited the same artifact.

Design Practice

Built the design function as a standing system — a team, a research practice embedded in every sprint, and explicit working agreements across Product, Engineering, and Data Science. Research wasn't a phase; it was an input. Critique cadences and shared quality standards kept four designers moving fast without diverging. Working agreements replaced ad-hoc coordination with predictable review cycles. Alongside the team-level work, I introduced the product catalog, OKR/KPI frameworks, and a Product Readiness Tracker that gave cross-functional teams shared language for planning and shipping.

Impact: Design moved from downstream consumer of strategy to a standing input at the leadership table. The frameworks were adopted across departments because they gave every function a common reference for prioritization and release.

Outcomes

• Aligned the company around a concrete product vision after months of abstract strategy discussion
• AU Labs became the primary tool for internal alignment, investor communication, and customer positioning
• Focused the roadmap on career mobility — the use case with multi-stakeholder value across employees, managers, and leaders
• Four-person design function operated across Product and Marketing without rework, sustained by shared quality standards and critique cadences instead of individual heroics
• Product catalog, OKR/KPI frameworks, and Product Readiness Tracker adopted across departments as the shared language for prioritization and release
• Design moved from downstream consumer of strategy to a standing input at the leadership table

The Unified Profile

What it is. A single-page view that collapsed a person's fragmented inputs — transcript, resume, work history — into one picture of skills, role-fit, gaps, and pathways. Not a dashboard with tabs. A single scrolling artifact ordered the way a decision-maker needs it: what can this person do above the fold; what could they do next below.

How it's built. Skills grouped by domain so the viewer pattern-matches against the role they're evaluating. Each skill carries proficiency and source-confidence indicators. Role-fit replaced a single match percentage — precise and wrong, communicating false confidence — with a visual match profile showing which skill clusters align and what the gap-closing pathway looks like. Progressive disclosure managed density: a 10-second scan layer, a deep-read layer, and an on-demand methodology layer, all served by the same artifact.

What it feeds. Investor presentations showed the platform's value without taxonomy diagrams. Sales demos replaced competitive-positioning decks with the profile's side-by-side comparison — the data arguing for itself. Product used it to align engineering on what to build next. The profile also settled the wedge debate: career mobility won because the profile was the career mobility tool, not a component of it.

What I'd do differently

• Killed the skills heatmap earlier. It had internal momentum and visual appeal, neither of which was a reason to build around it. I let the exploration run longer than the signal warranted.
• Made the hybrid design–product role's authority boundaries explicit from day one. The overlap was right; the ambiguity cost time.
• Delegated execution sooner. I held onto craft longer than the situation warranted, at the cost of the alignment work only I could do.