↖ ALL PROJECTS
Institutional Client Portal
Morgan Stanley at Work
[
SUMMARY
]
When I took over ICP, it was a set of agency-inherited concepts—Digital Vault, plan switching, calendars, alerts—each covering only happy paths with no shared logic across teams. Leading design end-to-end, I partnered with PMs, engineering, and compliance to map real error and permission states, pivot inherited concepts against user and compliance evidence, and build a reusable, accessible design-system foundation. What had been disconnected, single-purpose features became a unified, scalable enterprise portal ready for a 10-client pilot.
[
FEATURE 1
]
Uploading Multiple Files Without Losing Track of What Failed
When a plan sponsor uploads several documents at once, each file validates and uploads independently — one file can still be uploading while another has already failed or finished. Instead of a single progress bar that can't reflect a partial failure, each file shows its own status inline, and a failed upload (like a duplicate filename) surfaces a clear choice to replace it or keep the existing version.
[
FEATURE 2
]
Comments and Version History, Attached to the File Itself
Plan sponsors and advisors needed to discuss a document and track how it changed over time, without leaving the file or losing context. Opening a document surfaces a collapsible side panel with threaded comments and a version history log, so feedback and file changes live in the same place as the document instead of scattered across email threads.
[
FEATURE 3
]
Turning Open-Ended Events Into Structured, Compliant Tasks
Plan sponsors needed a simple way to see what task was due and when, without scheduling something that reads like an open calendar invite. Creating a task means picking a task category, selecting from a fixed set of topic tags, and setting a date, time, and time zone — there's no free-form title, description, or location field, since those needed to stay within compliance requirements.
[
MOST MEMORABLE MOMENT
]
I Don't Execute Blindly. I Execute Deliberately.
I was handed a direction that had already been decided, approved years before I joined. My job was to build it. But building without checking it never felt like the job to me. The plan was event-based, something like a shared calendar with meeting details, inherited from an old agency concept. By the time it reached me, it read less like a proposal and more like an instruction. Before I started building, I went back to what advisors and the business were saying people needed: not events, just a clear task, a due date, and a topic. Compliance rules ruled out the open-ended fields the event model depended on. So instead of building what was handed to me, I used that evidence to argue for a structured task model and worked with the team to get there. Being told what to build isn't the same as being told what's right.
[
PROJECT INFO
]
MY ROLE
Led early Digital Vault desk research beginning January 2022, then became lead designer for Digital Vault in March 2022 after the previous Senior Designer's departure. Owned the MVP design and delivery under a compressed timeline, then continued to lead post-MVP features, scalability work, design-system governance, and accessibility improvements.
TEAM
5 engineering squads (~4 developers + 1 QA each); 2 PMs during MVP, 3 PMs post-MVP; Design Manager (final design approval); Content Design; QA; LRC/Legal/Compliance; SMEs
DURATION
January 2022 – end of 2023 (~2 years)
TOOLS
Figma, FigJam, React Storybook, Jira, iOS VoiceOver, Adobe Analytics
[
IMPACT
]
Digital Vault shipped as the MVP in July 2022 despite an already-delayed timeline and a mid-project handoff from the previous designer, then passed governance review to launch a 10-client pilot two months later. Pilot users found the product easy to use and said it reduced how much they relied on email to exchange documents, and an early pilot survey returned an estimated NPS of around 66. In its first weeks, the pilot saw roughly five active users upload more than 100 files. Beyond the pilot, the underlying patterns — permission visibility, error recovery, multi-file states, and the comments/version-history system — became reusable enterprise workflow foundations, supporting an estimated five product lines and six products or platforms across the institution. The product was also shown at an internal global technology showcase and supported pre-sales conversations representing an estimated $200M+ in enterprise client scale.
Building a Self-Service Portal for Retirement Plan Sponsors Who'd Only Had Email and Manual Processes
Nobody Owned It. So I Did.
Ownership mindset
Dealing with ambiguity
Cross-functional collaboration
When this project started, the senior designer who owned Digital Vault gave two weeks' notice, and I was still junior, only six months into the company, in a domain I didn't know well. I stepped into the gap, took over the most complex remaining MVP feature, worked with two PMs to realign the scope and success criteria, and rebuilt a plan around a July deadline that had already slipped once. Digital Vault shipped to production on time, and the ten-client pilot launched two months later. I learned that ownership isn't assigned to you by title, you take it when the work needs someone to hold it.
Pretty Is Not Honest.
Data-driven design decisions
Feasibility assessment
Design decision articulation
Once the direction became clear for multi-file upload, the team considered an overall percentage bar to show progress, since it looked more polished than a plain list of statuses. I pushed back, since the API couldn't return a reliable percentage and a failed file would make an aggregate number misleading, and I proposed keeping the feedback at the file level: uploading, complete, or failed, even though it looked less impressive than a single moving bar. We removed the percentage and shipped truthful, file-level status instead. I learned that when a system can't support what it claims, the honest interface is the one worth defending.
Unknown Is My Invitation.
Accessibility (WCAG)
Curiosity & continuous learning
Inclusive design
As accessibility work picked up in late 2022, I found that our username and password hint text didn't work for screen readers, since reading every visible word made the login experience long and disorienting for non-visual users. No other internal product had solved this, so there was no pattern to copy, and I researched announcement order and necessary criteria myself, then partnered with content design on wording and with developers on a feasible structure. The new login pattern reached production and became an early internal reference for accessible login. I learned that when nothing exists to copy, the work is doable: it starts with research instead of a template.
