Try clicking around for a surprise!
AI Billing Codes
Lead Product Designer
Playback Health, 2025
Designed AI-powered billing codes that reduced documentation time by 40% while maintaining clinical accuracy.
Trial Sign-Up
Lead Product Designer
Playback Health, 2025
Reduced steps in sign-up flow that increased conversions by 35% and increased sign-ups by 8%.
Playback Health Website
Design & Development Lead, 2025
Playback Health, 2025
Redesigned marketing website to improve feature clarity and drive qualified trial sign-ups.
AI Ambient Notes
Founding Product Designer
Playback Health, 2023
Led 0→1 design of AI medical scribe that became the company's highest-revenue product with daily clinical use.
Marriott Bonvoy
Experience Designer
Code and Theory, 2023
Designed and delivered a component library to enable scalable promotional content across Marriott's digital platforms.
Aveeno for Kids Website
Experience Designer
Code and Theory, 2023
Redesigned dual-audience website that balanced parent education and advertising with child-friendly engagement.
Enyay Studio Website
Designed and built a gallery website for my ceramics business.
Microsoft for Educators
Translated user research into role-oriented experiences for Microsoft for Educators.
Bare Necessities
Audited a DTC website to create a more successful purchasing funnel and improve information architecture.
Mars Inc, Design System
Created a scalable, brand-compliant design system for Mars Inc. sites and delivered to their engineering team.
AI Billing Code Feature
Lead Product Designer, 2025
Led design of AI-powered billing code feature that reduced documentation time by 40% and decreased coding errors by 23% while building clinician trust in non-deterministic AI.
The billing LLM's non-deterministic nature required design solutions that built trust through transparency: visible confidence scores, transcript references, and explicit user confirmation.
AI Billing Code Feature · Phase 1
Gather technical requirements for billing codes and LLM
- Develop technical understanding and translate for product flow
- Preserve familiar UX for billing codes
- Align with technical constraints and AI LLM capabilities
- Consider user opinions on non-deterministic AI in designs
Understanding how our AI LLM generates billing codes
Partnered with engineering to understand how the domain-specific LLM generates CPT and ICD-10 codes, translating technical constraints into user-centered design requirements.
Addressing user skepticism of non-deterministic AI
- Design for non-deterministic AI output and integrate human-in-the-loop design
- Ensure user has full control over final output and provide opportunities for feedback and changes
- Communicate clearly with user via UX writing to ease users
Flow Structure
AI analyzes transcript
for recommended
CPT & ICD-10 codes
Technical LLM generates
codes with confidence levels
and orders them by relevance
(High, Medium, Low)
For each code:
• Code (e.g., "Z86.718")
• Natural language explanation
• Transcript reference
User reviews list and
manually confirms which codes
to send to EMR or copy
Understanding User Needs
Through user interviews with clinicians, I identified key tensions between AI assistance and billing accuracy that shaped the feature's human-in-the-loop design.
- Users have many codes memorized, so accuracy is very important.
- If user burden is too high, clinicians will not use feature.
- Users are often skeptical of AI, so UX needs to build trust.
- Users note process as “fatiguing” and “repetitive” so improvement > perfection.
- Users want to see the code before they submit to EMR.
Impact of inaccurate coding:
- Increased claim denials and rejections
- Delayed reimbursement cycles
- Higher administrative overhead for corrections and resubmissions
- Direct revenue loss from unbilled or underbilled services
AI Billing Code Feature · Phase 2
Iterate on design to align with technical requirements and business priorities
- Create wireframes that include all required info
- Preserve familiar UX to prevent user confusion
- Build trust for AI-generated content by adding "reasoning" to UI/UX
- Ensure every iteration reveals a piece of the design
Layout and Content Blocks
Explored multiple layouts to display AI-generated codes without disrupting familiar note-taking workflows.
Previous design
Layout of basic zones
Try existing components + carousel.
Nope, hides new functions
Wireframes
Iterated on vertical stacking to preserve existing functionality while introducing new AI features. Each wireframe tested how to surface confidence levels and transcript context without overwhelming users.
Considering existing functionality.
To preserve this → stack vertically
Using iteration discoveries to progress the wireframe
Adding in known components +
content to establish constraints
Finalizing Designs
Balanced design polish with development timelines, prioritizing clear feedback mechanisms over visual complexity. Positioned user controls prominently to reinforce clinician authority over AI suggestions.
Bottom toolbar added in.
Only one component left!
We chose a simple design to meet
development & delivery deadlines
Added the feedback section below
content to avoid distraction
AI Billing Code Feature · Phase 3
Deliver pixel perfect designs, design for contingencies, update flows and finish up
- Unify web and native UI/UX for company cohesion
- Create a scalable, comprehensive design library
- Build trust with inclusive, intelligent UX writing
Copy to Clipboard
Designed clipboard functionality with clear visual feedback and code formatting that matches EMR input requirements, reducing manual reformatting.
Sharing via Email
Created email sharing flow for practices with medical assistants who handle billing, allowing clinicians to delegate while maintaining oversight of AI-generated codes.
Submit to EMR + Confirmation
Designed direct EMR integration with explicit confirmation states to build trust. Clear success feedback reassures clinicians that codes were accurately submitted to their billing system.
Outcomes
The AI Billing Codes feature transformed tedious administrative work into a streamlined workflow that reduced documentation burden while improving accuracy. Through close collaboration between design, product, and engineering, I prioritized clinician trust and control by balancing technical AI capabilities with clinical workflow needs.
This human-in-the-loop approach proved critical to achieving high adoption and measurable practice-level impact.
Trial Sign-Up
Lead Product Designer, 2025
Redesigned the trial sign-up experience for Playback Health's domain-specific AI medical scribe, resulting in a 35% increase in conversions and 8% increase in user sign-ups.
This project was a close collaboration between Product Management and Product Design. What began as funnel drop-offs evolved into enhanced workflows, resulting in a 35% increase in conversions and 8% increase in user sign-ups.
Trial Sign-Up · Phase 1
Establish business needs and collect product data
- Understand user needs
- Identify conversion hurdles within sign-up experience
- Compare existing sign-up flow with best practices
- Define necessary data to be collected for sign-up
Subscription Funnel
Website
User views marketing page with CTA
Sign-Up
7-step flow that collects user info
Activation
User creates at least one note
Subscription
User pays a monthly/annual subscription
Understanding User Needs
User research revealed that clinicians needed to experience product value within their 14-day trial before committing. Any friction in sign-up directly prevented product adoption.
- User must justify the cost based on perceived value over 14 days.
- Users are short on time, highly intelligent, not always tech savvy.
- User must see value in the product before bids for subscription.
- User burdens should be deferred until after users see product value.
Competitor Analysis and Internal Audit
Competitive analysis of three medical scribe platforms revealed Playback required more steps than necessary—competitors collected similar data in fewer actions.
Competitor A
4 user actions
3 pages
8 data points
Competitor B
6 user actions
6 pages
12 data points
Competitor C
8 user actions
6 pages
10 data points
Playback
7 user actions
6 pages
8 data points
Proposed Changes
Consolidate the flow from 7 steps to 3 essential actions, eliminating redundant requests while preserving data needed for trial personalization.
Old Sign-Up Flow
Longer form
More steps
Less useful data collected
New Sign-Up Flow
Fewer steps
More relevant data collected
Lower user burden
Trial Sign-Up · Phase 2
Translate into UX/UI information architecture and user flows
- Gather needs from Marketing, Sales, Customer Success to deliver useful automation.
- Define main 'Trial Sign-Up' flow, considering user types (B2B, B2C, B2B2C)
- Simplify sign-up to minimize user friction
Detailed Sign-Up Flow
Original flow failed by front-loading security steps (MFA, password) before users experienced product value. The 7-step process created abandonment points at verification stages, contradicting clinicians' need for immediate trial access.
Practice Name
Need this to personalize templates
(phone or email)
Key
Simplified Sign-Up Flow
Streamlined flow to 3 core steps: combined name/email input, specialty selection, and trial welcome—reducing friction while maintaining personalization.
Specialties
Key
Trial Sign-Up · Phase 3
Deliver a seamless, intuitive user experience
- Unify web and native UI/UX for company cohesion
- Create a scalable, comprehensive design library
- Build trust with inclusive, intelligent UX writing
Previous Designs
The inconsistent UI between marketing site, sign-up, and product created confusion and failed to establish clear product value.
Marketing Website
Old Playback Notes Sign-Up
Playback Notes Product
UI Updates
Unified visual design and UX writing across touchpoints to create a cohesive experience from marketing to product.
Marketing Website
New Sign-Up Flow landing page
Playback Notes Product
Fully Responsive Design Library
Built a comprehensive design system with components, states, and styles to ensure consistent brand experience across web and native platforms.
Final Flow & Outcomes
The final flow shown below delivers a frictionless experience with consistent UI and clear guidance, welcoming clinicians directly into their trial.
About
Hi, I'm Nina, a product designer who focuses on AI-powered tools and 0→1 product development.
I'm currently lead product designer at Playback Health, where I own the design and direction of our clinical AI tools. I joined as a founding designer in 2023 and my work centers on translating complex LLM capabilities into interfaces that earn user trust, from ambient documentation to billing tools that help reduce administrative burden.
I studied architecture at Brown and RISD, where I learned to balance storytelling with technical requirements. That training shapes how I approach product work today: building systems that scale but still feel human.
Originally from Northern California, I've spent almost 5 years designing experiences that bridge technical complexity and real user needs. Always happy to connect.