I used my own Expert Roundtable tool to audit this portfolio site. The panel — Sarah Drasner, John Maeda, and Kelsey Hightower — produced a prioritized list of improvements across three tiers: quick wins, structural changes, and long-term polish.
The most interesting finding: the panel unanimously flagged that leading with a resume page signals "I'm looking for a job" rather than "look at what I built." The fix was to lead with work — a gallery of generated ontologies that demonstrates the system's output at scale. Credentials become supporting evidence, not the headline.
Other highlights from the deliberation:
The panel identified "chatbot vibes" as the single biggest perception risk. Streaming markdown during model generation pattern-matches to ChatGPT in visitors' minds, regardless of what the dashboard looks like after. The recommendation: never show the stream. Show the instrument panel with placeholders that populate.
On dark mode: the dashboard was designed for dark. Defaulting to light theme on first visit undermined the aesthetic. The panel recommended dark-first with an explicit toggle for users who prefer light.
On the model library: 60+ reference models is impressive evidence, but it was mixed into the homepage alongside the portfolio narrative. The panel recommended separating concerns — the homepage tells the story, a dedicated gallery page is the catalog.
Every recommendation traced to a specific acceptance criterion with a verifiable done-when condition. The full audit produced 10 tasks across three priority tiers, all executed in a single session.