Product Design
FinTech
Design Systems

My job was to make that model visible to clients and give the team a design foundation.
I joined Onramp as the first designer before its Series A as its first product designer, on a three-month contract.
I worked with the Head of Product and the engineering lead, and owned design across the product: the design system, money movement and team governance.
Role
Founding Product designer
Team
Head of Product
Engineering lead
Timeline
Jun – Aug 2025
Type
Contract
Pre-Series A startup
Outcome
Design system and core flows, handed off to engineering
Every withdrawal passes through several people and institutions. That’s what protects the client. But if they can’t see the steps, a secure process just feels slow.
Multi-institution custody set Onramp apart from exchanges, yet the product looked like any other crypto app.
Few products had ever shown a client a group of institutions signing a transaction.
Engineering shipped in one-week sprints. Without one source of truth, engineers had to ask which design was final.
Before designing new flows, I audited existing screens to find repeated and inconsistent patterns.
I built a tokenized library of type, icons, buttons, inputs, tables, vault cards and a details side panel.

The audit: every pattern in the product, reduced to one set of parts.
Showing the whole path up front. The request page lists every step before the client starts, so later waits feel expected, not broken.
Keeping verification in the flow. The client books the video call inside the withdrawal, not in a separate email thread, so the request never leaves the product.
Naming the institutions. The last step shows each institution, its role and whether it has signed.
Letting 2-of-3 explain itself. The backup signer stays on screen as “Not signing”, so clients learn that two keys are enough without a tutorial.
Locking Submit until the quorum signs. The button only becomes active once two keys have signed.


Organizations put several people on one account, so they needed maker/checker: one person requests, another approves.
Instead of exposing quorum logic, I framed it as three named policies, each with its trade-off spelled out.
Single approver (1 of N)
Another Admin or Signer approves sensitive actions.
Dual approver (2 of N)
Two different people approve. The most secure.
No policy
Users approve their own changes. Flagged as not recommended for teams.

The current policy, in one sentence a team can check.
A table shows exactly which actions need approval: Bitcoin transactions, USD and banking, the address book and user management. I refined it over three rounds of stakeholder feedback.
Policies decide how many people must agree. Roles decide who can act. I designed a roles matrix (View-only, Initiator, Spender, Admin, Owner) that sets who can view, start, spend or administer funds.
In three months, I designed a component library and the withdrawal, approval and account settings flows, then handed them off to engineering.
Engineers built from one library and a final version in Figma.
Clients see who is protecting their money at every step of a withdrawal.
Organizations choose an approval policy and roles in plain language, with the risks called out.
Lending, IRAs and new custody partners can build on the same patterns.
In financial products, a step the client can’t see feels like a delay. A step they can see feels like protection.
Spending the first weeks on an audit and a library felt slow, but it’s why the later flows came together quickly and consistently.