← All work
Galileo Financial Technologies (SoFi)

What it takes for operators to trust a system that moves $26B a year

Galileo’s Console is the control layer for some of the biggest names in fintech — SoFi, Chime, Robinhood, Monzo, Dave. When I arrived, the operators configuring card programs and running settlement were drowning in a legacy interface that demanded constant manual intervention. I treated the redesign as an experiment in a question I’d been circling for years: can certainty in a complex system be designed deliberately, rather than accumulated by accident? The evidence came back rather clearly.

Role Principal Product Designer
Timeline July 2021 – February 2023
Team Led 2 contractors + design agency
Scope End-to-end Console redesign
The question

Can operator certainty in a system moving $26B a year be designed deliberately — or is it only ever earned slowly, by accident?

My role

Principal designer. End-to-end Console redesign, information architecture to pixel — plus the design system and handoff practice to sustain it. Led two contractors and an agency.

The evidence

90% less manual intervention, 65% fewer implementation bugs, onboarding time halved, NPS +2 — across clients at SoFi, Chime, and Robinhood scale.

90%
Less manual
intervention
65%
Fewer implementation
bugs
+2
NPS score
increase
50%
Faster
onboarding
The Question

Operators had no tolerance for ambiguity about where their money was. The platform gave them nothing but ambiguity.

Galileo processes over $26B in annual transactions for fintechs building on its infrastructure. But the operator-facing Console — where clients configure card programs, set authorization limits, manage payouts, and monitor settlement — was built for functionality, not usability.

The result: operations teams couldn't configure programs without calling support. Manual intervention was required for workflows that should have been self-service. The interface patterns were inconsistent across financial domains, and years of rapid growth had created deep technical debt in the UX layer.

Meanwhile, Galileo's clients — SoFi, Chime, Robinhood — were growing fast and demanding a platform experience that matched the quality of the APIs they'd built on.

"The operators I shadowed weren't confused by the financial concepts. They were confused by the interface. They knew exactly what they needed to configure — they just couldn't find it, couldn't trust what they were seeing, and couldn't tell if their changes had taken effect."
— From discovery research, August 2021
The Approach

I redesigned the Console from information architecture to pixel — and built the design function to sustain it.

Three working hypotheses: that the Console could become genuinely self-service (what we called “DIYability”); that a design system could scale across financial domains; and that structured handoff with engineering would end the pattern of shipped-but-broken features. All three held.

Fig. 01 — The configuration spine: the object model underneath every Console workflow
Program
→
Products
→
Parameters
→
Approval state
→
Audit trail
Mapping this spine came before any UI: nothing touches money movement without traversing an approval state and leaving an audit entry.
001

Parameter configuration

Redesigned how operators manage card authorization limits, fees, transaction rules, and product settings. Introduced intuitive controls for complex financial parameters, a compare feature for configuration versions, and clear confirmation states so operators always knew what changed.

002

Role-based access control

Designed a comprehensive permissions system with separation of duties — letting administrators assign granular access by workspace and role. Critical for compliance in a platform where a misconfiguration can affect millions of dollars in transactions.

003

Approval flows and audit history

Built multi-stage approval workflows for configuration changes, with full audit trails. Operators could now trace who changed what, when, and why — turning a compliance liability into a confidence signal.

004

Enterprise design system

Spearheaded development of a comprehensive design system — grid systems, color palette, typography, and a component library with specialized financial interface patterns. Created governance and documentation that let the system scale across teams.

Trade-offs — What I Chose Not to Build

The discipline was in what we refused to build.

Honest timestamps over fake real-time

Engineering could not guarantee sub-minute freshness on settlement data. Rather than imply live figures, every number carries an explicit “as of” timestamp. It looks slower. It is more trustworthy — and operators told us so.

Friction, kept deliberately

Product pushed for one-click configuration changes. We kept multi-stage confirmation for anything touching money movement: an operator moving millions should never wonder whether a click “took.” Speed lost the argument to certainty.

Personalized dashboards, cut

A customizable home screen was scoped for v1. We cut it to fund approval flows and audit history instead. Configurable widgets delight demos; audit trails retain clients.

Results

The results were unambiguous: the redesign changed how Galileo’s clients operated.

Operational efficiency

90% reduction in manual intervention in payment configuration and settlement workflows. Transaction throughput capacity improved 40%. Operators could self-serve configurations that previously required a support call.

Development quality

65% fewer implementation bugs through structured design-to-development workflows. Design-to-dev handoff issues dropped 50%. New feature delivery accelerated 35%.

Operator experience

NPS increased by 2 full points. New platform onboarding time cut in half. Support tickets for interface issues dropped 70%.

Business impact

Supported scaling to $26B+ in annual transaction volume. Enabled expansion into new financial product categories. Strengthened Galileo's competitive position as the BaaS infrastructure standard.

"Operators in financial infrastructure don't want delight; they want certainty. Every decision came back to one question — does the operator know exactly what just happened? I have since found the same question sitting at the center of AI interfaces. That transfer is the most useful thing Galileo taught me."
— What carried forward

Working on a system where operator trust is the constraint?

I do this work fractionally — embedded with your team, shipping real screens.

Get in touch