Self-initiated
Enterprise Product
Role
Everything, on my own
Disciplines
Product Design, Systems, Prototyping
Client
Self-initiated
Platform
Web, desktop first, in English
01
Vovini is a web app for juniors who get feedback in 1:1s and reviews. You write down what was said, Vovini proposes what it was about, and when the same note comes back it says so and asks what you are going to do differently before the next 1:1.
I designed it end to end on my own: the brief and a log of every decision, a design system, the flows, the data model and an architecture proposal, every screen from the landing page to settings, and a working prototype with motion.
It comes from supporting a Head of Design directly: monthly 1:1s, work reviews, and mentoring a junior designer who started without much confidence and became one of the strongest people on the team.
02
Feedback doesn't get forgotten, it gets repeated
The feedback that matters in a 1:1 is usually heard. It just lives in memory. Notes end up in a notebook, a Slack draft or nowhere, so there is nothing to hold them against, and six months later the same note arrives and feels new.
It fails in order. Nobody notices a repeat they never wrote down, and nobody acts on a pattern they never noticed. So the product had to answer remembering, noticing and acting, in that order.
The tools that exist for this mostly sit on the other side of the table: review cycles, performance platforms, dashboards for managers. What someone writes down after a 1:1 is honest in a way a shared document never is, so this one had to belong to the person hearing the feedback.
03
The system before the screens, and every option on one board
No screen was drawn until everything it depended on was on a board. The design system came first, taken from four reference products for the feel: airy, one purple carrying the brand, pastel surfaces that mean something. Then a working board with the loop, the data model, the options still open and the sitemap.
Decisions went into a written brief as they were settled, and anything still open went into a backlog instead of blocking the work. Three rules came out of the system and held on every screen: category colour says which, chart colour says how much, and amber only ever means you have heard this before.
The prototype runs the whole loop in a browser with example data, from the landing page to the 1:1. The screens on this page are captured from it, not redrawn beside it.
The working board: the loop, the data model, the sitemap and parts of the design system, before any screen existed.
04
The Tension
What I chose
What it cost
The Tension
What I chose
What it cost
The Tension
What I chose
What it cost
The Tension
What I chose
What it cost
The Tension
What I chose
What it cost
05
A note, a reading, and one question
Using Vovini starts right after a 1:1. The calendar already knows who it was with and when, so the page holds only the note: pasted in, or written the way you would tell a friend.
Vovini reads it once and shows its working: the categories, their weights and the sentence behind each one. When the note repeats something heard before, an amber callout offers two ways on, act on it now or open the thread.
Patterns shows how much each category came up month by month, and the threads opened along the way. On the morning of the next 1:1, Reviews brings back what was promised. Everything works in light and dark, and the prototype below runs the whole loop.
06
HOW IT WAS REVIEWED
WHAT SHIPPED
07
Nobody has logged a real 1:1 in it yet
What exists is every screen from the landing page to settings, a prototype that runs the whole loop with example data, a design system, and an architecture proposal. No junior has used it with a real 1:1.
The first question comes before any pattern: whether people write anything down after a 1:1 at all. Capture is the front door, and if notes do not arrive there is nothing for Vovini to read and nothing else is worth building yet.
Testing. Not run yet. The first round would ask a few juniors to log their next two 1:1s, then talk through whether the weights read as a guess or a verdict, and whether not yet felt private enough to answer honestly.
Built so far. Every screen, in light and dark. A prototype of the whole loop. A design system, the flows, the data model and an architecture proposal.
Still open. Who writes the commitment. What happens when a commitment misses two 1:1s in a row. Whether six fixed categories survive contact with real notes.
08
The system came first, and it made saying no easier
Early on, the reference products I had chosen for the feel were being treated as content to describe rather than a look to build from. I stopped the work there and had the design system built first, then a board with every option still open, and only then any screens. I settled most later arguments by pointing at a rule.
This was built with Claude as a collaborator, the same way root was. It wrote the prototype, the design canvas and first drafts of most of the copy, proposed the six categories and the architecture, and generated the photos and illustrations. The people in those images are not real.
My part was deciding. That the app belongs to the junior. That an aha has to ask for an action. That the review is the monthly 1:1 from a real calendar, that a single note can open a commitment, and that a thread’s title can be rewritten. And the name, twice.