Projects & Case Studies | Prashant Arya

01

Ambria

ICU nurses shouldn't have to manually measure what machines can monitor continuously.

Penn Medicine Co-Founder + Product Lead Medical Device 0 to 1

The Problem

Urine output is one of the most critical signals in an ICU, a leading indicator of shock, kidney failure, and fluid imbalance. Yet in most hospitals, nurses still measure it manually, every hour, for every catheterized patient. The process is inaccurate, biohazardous, and occupies time that should be spent on higher-acuity care. Physicians make decisions on data they know is unreliable. The gap between what's clinically needed and what the workflow delivers is enormous.

My Role

Co-Founder and Product Lead, owning the full product lifecycle from clinical discovery through IP transfer. Drove user research, product strategy, requirements definition, iterative design, and the GTM and business case for Penn Medicine's innovation pipeline.

How I Built It

Embedded in Penn Medicine MICU wards to observe the workflow firsthand. The problem was far more systemic than initially framed by clinical staff Interviewed nurses, intensivists, and hospital administrators to map pain across the full care team, not just the bedside Identified that the core failure was the measurement-disposal loop: inaccurate data entry, biohazard exposure, and EHR fragmentation all stemmed from the same broken process Designed an integrated device with automated continuous measurement, spill-free disposal, one-touch bag removal, direct EMR connectivity, and configurable clinical alarms Created go-to-market strategy and business case anchored to hard savings per ICU deployment

Outcome

80% reduction in urine disposal time. $98,100 in annual savings per ICU. All IP transferred to University of Pennsylvania. Ambria is currently in refinement at Penn Medicine Innovation Center on the FDA approval pathway.

02

Duolyfe

The productivity market solved for doing more. Nobody solved for stopping.

Founder Consumer App 0 to 1 Behavioral Design

The Problem

COVID didn't just move work home. It eliminated every environmental cue that told people to stop working. The commute, the colleague leaving, the physical context switch: gone. 100,000+ productivity apps existed to help people work more efficiently. None existed to help them stop. The market had built entirely in one direction, and the behavioral gap it left behind was real and measurable.

My Role

Founder. Set the product vision, led discovery, defined the behavioral model, ran design, and coordinated a two-engineer team toward an MVP. Responsible for every strategic call from hypothesis to prototype.

How I Built It

15 user interviews and 100+ surveys, not to validate the problem (which was obvious) but to understand the mechanism: why don't people stop even when they want to? Core insight: it wasn't a time problem or a willpower problem. It was a motivation vacuum: people lacked an intrinsic reason to invest in their personal time Reframed the design challenge: don't block work, create pull toward life. The solution needed to generate its own motivation, not borrow from willpower 5 design iterations from paper prototypes to high-fidelity Figma, testing engagement models across habit tracking, social accountability, and companion mechanics Applied RICE framework to scope MVP: prioritized the Buddy mechanic above all others as the highest-engagement, lowest-complexity path to the core behavior change

Outcome

Shipped a virtual companion, the "Buddy", whose wellbeing mirrors the user's own self-care. Take a walk, your buddy thrives. Skip dinner, it shows. Built extrinsic accountability that creates intrinsic habit change. MVP in development for Play Store launch.

03

Redesigning Amazon Rufus

AI that recommends without reasoning doesn't build confidence. It transfers the work.

Georgia Tech MS AI HCI Research AI Trust Design Conversational UX

View Full Report

The Problem

Amazon Rufus returns a shortlist. It doesn't explain the shortlist. Users can't determine whether results are ranked by reviews, by margin, or by algorithm, so the AI creates a new decision to make rather than eliminating one. The failure isn't the technology. It's a product design choice to omit reasoning, and it's costing Rufus the trust that would make it actually useful.

My Role

Researcher and Product Designer, owning the full two-iteration HCI design lifecycle. Set the research hypothesis, designed and evaluated three competing prototypes, synthesized findings into a high-fidelity solution, and documented the design rationale at publication quality.

How I Built It

Structured needfinding: user surveys on shopping behavior plus a rigorous heuristic evaluation of the live Rufus interface against Nielsen's 10 usability principles Framed three distinct design hypotheses (transparency in output, constraint-setting before results, guided comparison after results) and built a low-fi prototype for each Ran structured evaluations with 18 participants to isolate which intervention had the highest impact on perceived trust and decision confidence Synthesized across all three into a single high-fidelity design: transparent recommendation cards with visible ranking criteria, a lightweight pre-search intake form, and an optional guided decision flow for high-consideration purchases

Outcome

A high-fidelity prototype that makes AI reasoning visible at the point of recommendation, reducing the trust gap that drives abandonment and turning Rufus from a list-generator into a genuine decision-support tool.

Read the full report