Skip to main content

Finominal

Redesigned portfolio analysis workflows for advisors. Built and shipped features end-to-end using Claude Code.

About the Project

Finominal is built for professional investors like Portfolio Managers, Advisors, Distributors, and Family Offices. Existing tools either oversimplify or overwhelm. Finominal tries to sit between both.

My Role

Sole Product Designer END TO END

Owned the experience from strategy to shipping. Research, IA, interaction design, and QA, with no other designer on the team.

Vibe Coded CLAUDE CODE

Skipped the usual design-to-dev handoff. Iterated directly against production data and shipped UI changes myself using Claude Code.

The Problem

When I joined, Finominal was live with paying clients and carrying significant design debt. The product had grown feature-first. Data existed but decision support didn’t.

Problem 01

No information architecture

Problem 02

Every data point, equal weight

Problem 03

Comparison required five screens

The debt was not in the code. It was in the gap between what advisors needed to decide and what the screens let them see.

My Approach

01

AUDIT

Map every flow

I mapped every flow, screen by screen. Documented clicks, dead ends, and repeated patterns. No IA existed so building one was the first move.

02

SIGNAL FILTERING

Separate noise from signal

Feedback came through Slack, mostly sales demo notes. I filtered for workflow failures, not feature requests. "Can we add X?" was noise. "They opened Excel to do Y" was signal.

03

ITERATE

Redesign, repeat

Audit, identify root cause, redesign, repeat. Decisions made directly with the founder.

How might we make complex portfolio analysis feel effortless for advisors, without stripping the depth they depend on?

This meant reducing friction in portfolio creation and analysis workflows, presenting data in ways that aid interpretation, and restructuring the IA so the right data surfaces at the right time.

01Fund Explorer

Users were struggling to find what they needed. Navigation complaints were coming in, and we were already moving to a new design system. That combination made a full redesign worth doing rather than patching.

The Problem

Problem 01

Filters don’t scale

Problem 02

No column hierarchy

Problem 03

Tools with no entry point

WHAT CHANGED

1.Sidebar collapses. Full table width restored when filters aren’t in use.
2.First three columns fixed. Context stays while scrolling right.
3.Watchlist tab added. Same layout as Fund Explorer. One pattern, two views.
4.Actions moved above the table. Always accessible, never covering rows.

02Dashboard

The dashboard is where advisors keep their saved portfolios and send them into the tools. The redesign was already underway when grouping came in as a new feature. Portfolios had piled up in one flat list, and the actions above it blended into the background.

The Problem

Problem 01

Actions went unnoticed

Problem 02

No way to group portfolios

Problem 03

Actions with no hierarchy

Every version ran into the same question: where do the existing portfolios go? A Default folder answered it. Nothing moves until the user moves it.

WHAT CHANGED

1.Folders moved to a second side rail. Every folder is always in view, one click to switch, and the rail collapses when the table needs room.
2.Existing portfolios land in a Default folder. Nothing is lost in the move, and users regroup them when they’re ready.
3.The table stays flat, one folder at a time, so sorting works again with no groups to break.
4.Search and Create Folder sit at the top of the rail. Finding a folder doesn’t depend on scrolling.

RESULT

More than 20 users started using folders in the first week.

03Graham AI

Graham AI answers questions about funds and portfolios using the platform’s own data. The answers already existed, spread across Fund Analyzer and Fund Explorer, but reaching one meant choosing a tool, finding the fund, and digging through tabs. No client asked for this.

The Problem

Problem 01

Answers spread across tools

Problem 02

Narrow questions, long paths

Path comparison

TODAY

Choose a toolFind ARKKOpen the detail pagePick the Performance tab

GRAHAM AI

“Analyze performance of $ARKK”Answer and chart

What I designed

1.What you can ask. Performance, comparisons, drawdowns, stress tests, portfolios.
2.How you point at things. $ for a fund, / for a portfolio, so Graham isn’t guessing.
3.What it won’t do. Give advice. Every answer shows its sources instead.

WHAT CHANGED

1.Answer and evidence split. The plain-language read sits on the left, and the charts on the right.
2.Charts became tabs, one at a time, each with a label and a one-line takeaway.
3.Related research, videos and a presentation sit under the answer, so the next step is in reach.

Where it stands

In internal testing. Some answers still come back messy, and we’re working out why.

Outcome

I rebuilt 8 tools around how advisors work, and built 3 MVPs and 4 features from scratch. One MVP reads an uploaded statement and maps the holdings. Another tells advisors whether a portfolio sits above or below their risk threshold. I set up Amplitude so we stopped guessing about usage.

Learnings

Shipping code yourself means you can’t design around ambiguity. Every edge case you didn’t think through becomes your problem to solve. Working without a PM taught me that being right about design doesn’t matter if it doesn’t move what the business actually needs.

Next case studyInvest ITAI trading insights, shipped live