Savvly

I was the sole designer at Savvly. This case study covers a project I led: redesigning the client dashboard.

Savvly is a longevity benefit: a financial product that pools contributions and pays out structured cash at ages 80, 85, 90, and 95. The promise is protection against outliving your savings. The existing experience told the wrong story.

The estimator led with projections: how much you could get, under which scenarios, compared to what. It framed Savvly as an investment competing on returns. But most people didn't even know longevity benefits existed. There was no competing product — there was no existing product at all.

User research confirmed it: retirement planning made people uncomfortable — few had named "outliving your savings" as a specific risk. So the experience couldn't open with a calculator; calculators are for people who've already decided to engage.

The framing: nobody buys home insurance for the returns — they buy it because their house might burn down. Savvly needed the same shift — from investment product to protection against a fear.

The estimator lives in the employee dashboard, where clients explore their benefit after onboarding. The redesign cut inputs, added progressive disclosure, and restructured the flow to establish the risk before the solution: what longevity risk is, why typical retirement planning misses it, then Savvly as the answer.

User testing showed people distrust financial projections. What landed was not "here's what you might get," but "here's what happens if you live to 95 and run out." I redrew the key visual as a shark-fin curve, which emphasized that payouts grow largest exactly when savings tend to run out.

Payout milestones became gift cards instead of graph points, arriving over time rather than one lump-sum projection.

The Shed

A transcription is learning a recorded solo by ear, note by note. One of the foundational ways jazz musicians learn is by copying the way the greats played. The process requires slowing audio down, looping short phrases, and marking moments to return to.

Most musicians have their own cobbled-together process: YouTube and a separate metronome, Chrome extensions, an MP3 download into whatever playback software they have. But across all methods, the workflow is fragmented and constantly pulls you out of the music.

I built this because my own process was fragmented — and I realized that only part of my slowness was my ear. The rest was the clumsy process itself.

I observed three jazz musicians transcribing in their natural workflow before designing anything.

• 
  One had opened a professional audio editing suite — not to edit anything, but purely to see the waveform.
• 
  Another was using his phone's YouTube app, double-tapping to skip forward and back, but overshooting each time.
• 
  A third kept switching apps mid-session to use a separate metronome.

I then ran think-aloud sessions with five musicians on an early version of the tool, and sent them iterations as I built.

Speed-relative seeking: At 0.5× speed, skipping 5 seconds forward feels like 10 real seconds of music, which felt too far. I made skip distance proportional to playback speed, so at half speed you move half as far.

Built-in metronome with BPM sync: Once I saw the app-switching happening, the fix was obvious. Tap or detect the tempo, and the metronome syncs to it. When you slow the playback down, the metronome slows with it. You can practice a phrase at full speed, drop it to half, and practice again without ever reentering the tempo.

Keyboard shortcuts: Think-aloud sessions made clear I needed more than YouTube's defaults. I added up/down arrows for speed and a key to drop a marker mid-phrase without breaking flow.

A-to-B looping: Set two points in the audio and loop between them indefinitely while you work out a phrase.

A single-page tool that consolidates the full transcription workflow:

All without ever leaving the browser.

An unexpected use case emerged after launch: composition arrangers using the tool, working through a recording phrase by phrase to transcribe a piece before arranging it for a new ensemble!

200+ musicians have used The Shed so far. A few responses since launch:

"The looping feature is probably the most visually interactive system I've dealt with. I finished my Sonny Stitt 16th note line really fast — I'll use this for every 16th note Bird line."

"Much easier than what I used to do — download a YouTube video, open it in Audacity."

"Just played around with it tonight and really dig it! Got some handy features I've been looking for."

EmpowerHealth

This project started within the Fung Fellowship at UC Berkeley, where I worked alongside Ishita Srivastava, Ashley Huerta, and Nessia Dambal. After the fellowship concluded, I continued developing the project as design lead with Ishita Srivastava and Jonanne Talebloo.

Chronic illness is invisible, non-linear, and hard to explain in fifteen minutes, which is often all a patient gets. By the time someone with a chronic condition sits down with a doctor, they're carrying years of context: misdiagnoses, fluctuating symptoms, physical and mental pain that are deeply entangled. And yet almost none of it makes it into the room. Doctors don't have access to the history, don't have time to absorb it, or don't believe it.

The result: patients feel unseen, and providers can't do better.

We interviewed users with chronic illness, and providers. Between them: years of misdiagnosis, $17,000 out of pocket, 50+ specialist appointments for one condition, a job lost mid-treatment.

Being dismissed or not believed: every patient named this. One had lived with chronic pain for years; doctors suspected fibromyalgia but refused to record it, fearing the label would make future providers dismiss any new symptom as "just your fibromyalgia." Without the diagnosis, he couldn't get treatment. With it, he'd lose credibility. The system was protecting itself at his expense.

The proof burden. Invisible illness means constantly justifying pain — to doctors, employers, family, friends. One patient scratched their own arm mid-appointment just to be believed.

Exhaustion from re-explaining. Starting over at every new appointment, every new specialist, with no record of what came before. "I want a central place where everything is in one place for all doctors, irrespective of the hospital."

The mind-body connection, discovered alone. Nearly every patient had privately figured out that stress triggered flares, that emotional and physical pain were entangled. Almost none had a doctor name it. Dr. Bennett put it plainly: stress is a bodily stressor that literally fires pain receptors. The system wasn't tracking it.

Isolation. One patient said they sometimes made up excuses rather than explain their illness. Another said "I never felt sick enough" — real symptoms, but excluded from the chronic illness community because their condition wasn't visibly debilitating enough.

Providers were just as constrained. Dr. Bennett described watching patients' lives get "smaller and smaller and smaller, and the focus on the pain get larger and larger" — a spiral he wanted to interrupt but had no infrastructure to address in a standard appointment.

How might we help patients and providers build a relationship where chronic illness is understood, not just managed?

We designed two linked portals: a patient-facing mobile app and a provider dashboard.

The patient app centered on a daily diary card — a design that holds space for whatever a user was carrying that day without making the process feel clinical or burdensome.

The provider dashboard used AI-summarized data from patient diary cards before appointments, flagging new symptoms, mood trends, and key concerns the patient had noted, so the fifteen minutes could be spent on care rather than catch-up.

The design was a first sketch, early-stage and untested with real patients or providers. Next steps would be usability testing, especially on the diary card interaction and how doctors want to scan summarized data under time pressure.

The fibromyalgia catch-22 alone points to a harder question the design doesn't yet answer: how do you build trust into a system when the system itself has already broken it?

Pathrise Education

Pathrise is a platform for college consultants to manage and support their students through the admissions process. My team designed the consultant-facing dashboard, giving consultants a centralized view of each student's tasks, deadlines, and application progress across their full caseload.

After sitting in on consultant interviews, I noticed a recurring gap: students showed up to writing sessions with no organized material. They managed essays across notes, docs, and spreadsheets, reused content across schools with no system, and tracked deadlines separately from writing progress. Consultants ended up doing logistics instead of guidance.

To find the underlying structure, I mapped my own senior-year application materials — stories, brainstorming docs, spreadsheets — until a repeatable, phase-by-phase formula emerged. From that, I proposed and designed a student-facing brainstorm portal.

The portal walks students through five phases: building a story bank, grouping experiences into themes, matching themes to essay types, drafting outlines, and generating a college-specific timeline. Instead of managing ten separate essays for ten schools, students build one narrative system that maps across every application.

After presenting, the team noted this workflow is most relevant for high-achieving students applying to many schools, which aligns well with Pathrise's existing consultant clientele.