Rebuilding an EHR with two-decade-old design as the only designer in the company.

Remedly was a solid EHR with a 2000s-looking product. The functionality was strong — practice management, charting, billing, patient records all worked. But the visual design dated back to when the CEO had built the original product himself, and the company had carried that styling forward year after year. By the time I joined, the product looked a generation behind what its users expected.

My job was to rebuild it — alone. I was the only designer in the company. A massive EHR with hundreds of pages, no design system, and no documented patterns. I owned the research and stakeholder interviews, the design system from zero, every page redesign, and every conversation with engineering to get the work shipped at pace.

We resolved it by going slowly first, then fast. Before any design work, I cataloged the entire application — every view, every button, every data point — so we wouldn't lose functionality in translation. Then I built a design system to replace the existing components. Then I worked the application page by page, with a daily review loop: design → medical consultant for HIPAA and clinical accuracy → CEO for product strategy → iterate → engineering → user verification.

The shipped product cut time per task by 60 to 70%. The minimalist visual language was chosen to age well — the kind of system that holds up for ten years rather than needing another redesign in three, which mattered for a 20–30-person company that couldn't afford to do this twice. Alongside the EHR, two more things shipped: a patient-facing mobile app with gamified intake and in-app communication, and a 2019 concept for voice-driven patient-visit capture that raised the funding round which kept the company going.

The constraint shaped the method. Remedly was a 20–30-person startup with a small user base; full quantitative UX research wasn't an option. So I went qualitative-heavy: interviewed existing users, the customer-experience and sales departments, clinical staff, and Remedly's clinical board members. n=22 across the customer-interview pool, plus internal stakeholders. The first deliverable was a full inventory of the existing application — every page, every button, every data point — paired with what each user role actually did with it.

The customer signals pointed in one direction. 78% of users disliked the UI; 69% said tasks took too long; 75% said the product didn't align with the sales pitch; and 72% used Remedly primarily for scheduling, not for the EHR features. Stack-ranked positives and negatives from the n=22 pool gave the work a target list — the body map and the calendar/booking flow were broadly liked; the click count, the popups, and the time to record medical history were the consistent complaints. 20 of 22 respondents named recording medical history as the worst friction in the original product.

Competitor benchmarking turned the complaints into a target. I rated Remedly against Epic and Cerner on four axes — medical content quality, workflow speed, feature breadth, and usability/UI. Remedly was behind on every one. From Epic, the lessons to copy were customization and a large-userbase UX vocabulary; from Cerner, left-menu navigation, auto-dictation, tables for readability, and a restrained color palette. From both, the failure modes to avoid: oversized icons, outdated visuals, popup-heavy workflows, and tab-based navigation that hides what users actually need.

1. CEHR certification. The single largest financial incentive and a baseline credibility requirement. 22 of 22 respondents flagged its absence.
2. UI rebuild. Analyze existing workflows, modernize visuals, and reduce click count systematically. The 78% UI-dislike and 69% task-too-long signals collapse into this priority.
3. Medical-forms customization. The largest feature gap, named by 20 of 22 respondents. The fix was a customization tool, not more pre-built templates.
4. Booking system rebuild. A new clinical-first system that fixes what 18 of 22 already liked about Remedly's calendar — without breaking what worked.
5. Billing functionality. The second-largest functional complaint (19 of 22). On-the-spot codes and macros to address smaller-clinic workflows.
6. Patient-facing app. The post-encounter time problem starts pre-encounter — confirm medical history before the patient walks in. This priority became the patient mobile-app project.

Pick a visual language that ages well. A 20–30-person company can't afford to rebuild its design every three years. I picked minimalism because it doesn't decay the way a trend-driven aesthetic does — done well, a minimalist system reads as current ten years later, the way minimalist work from the last fifty years still feels contemporary today. The financial argument was as important as the aesthetic one.

Optimize for time-per-task, not screens-per-flow. Public data on EHR usage at the time put clinical data entry at three times the patient visit length — 15 minutes of charting for every 5 minutes face-to-face. Every interaction in the new design was scrutinized against that ratio. Faster wasn't the goal in the abstract; faster without losing accuracy or HIPAA compliance was the goal in practice.

Daily collaboration as the engine. I would design a page, review the clinical and HIPAA implications with the medical consultant, review the workflow and product-strategy implications with the CEO, integrate the feedback, and have the next version ready by end of day. The CEO trusted the design decisions and didn't micromanage, which made the loop fast. Customer verification followed once the build was in front of users. Slow per page; fast across the project.

Time to record medical history dropped 43% — measured against the same workflow that 20 of 22 pre-research respondents had named the worst friction point in the original product. Across the redesigned task surface generally, time-per-task improved by an estimated 60–70% depending on the flow. The persona re-tests across three feature areas — Assessment & Plan, List View, and Body Map — flipped the per-feature ratings on UI, usability, speed, and content from mostly negative in the pre-update review to mostly positive in the post-update review.

The strategy was to copy what worked from the competitors and avoid what didn't. Implemented from Epic and Cerner: the left-menu navigation pattern, auto-dictation that turns interactions into notes, customization as a first-class capability, tables and structured sections for readability, and the ability to scroll through sections instead of clicking between them. Left in their products: cognitive overload from heavy color and icon use, busy workflows, lack of hierarchy, popup-driven customization, and overcrowded top menus. The post-update prototype was the first time Remedly's product looked like it belonged in the same category as Epic and Cerner instead of a generation behind.

Two design ideas were inventions, not imports. "Creation boxes" replaced popups during the encounter flow — a contextual interface that surfaces inline when a user clicks "add," with a "save & next" button so adding multiple items doesn't require reopening the same dialog. The split-screen body map became a collapsible widget that occupied 60% less screen space while preserving every interaction the previous product had supported, plus a muscle/skeletal toggle and a zoom view for focused encounters. On-the-spot billing codes with macros let smaller clinics close out a visit without leaving the encounter flow.

The voice-AI concept turned out to be the right bet, five years early. In 2019, ChatGPT and Gemini didn't exist. Image-to-image was the cutting edge of AI in design school. The idea of a clinical visit transcribed and structured by software in real time was genuinely speculative. The concept raised the company's funding round — and five years later, ambient-AI scribing became a core EHR feature category. The instinct was right; the timing was the constraint.

What I carried forward into the next roles. The discipline of designing for time-per-task. The conviction that a rebuild needs a catalog before a comp. The habit of a daily review loop with non-design stakeholders. And the lesson that for a small company, picking a visual language that ages well is a financial decision, not just an aesthetic one. All of these moved with me into Sr. Product Designer work at Science 37 and the Director of Product role at Lycia AI.