MVP, V1.0 SHIPPED

Creating an AI-driven early-stage cancer detection platform

As the sole designer, I collaborated closely with Dr. Liberty Foreman’s UK-based team, effectively navigating time zones to create a seamless user experience. My design approach focused on user-centered strategies tailored for AI-driven products, emphasizing real-time, understandable feedback based on AI results.

Team
1x Product Manager, 1x Product Designer, 4x Software developers
Platform
Desktop
Time
12 months, 2023-2024
Role
Sole Designer

overview

What is Sierra Medical

An AI-driven healthcare platform focused on simplifying early-stage cancer diagnostics with a swab test.

Objectives

Launch MVP in 8 months to acquire early adopters

The primary objective of this project is to Launch MVP in 8 months to acquire early adopters (private clinics and labs). To achieve the goals, I meet with stakeholder and set strategic objectives to support the product building from scratch.

scope

Building the basics and scaling up

  • Streamline cancer test experience.
  • Provide good user experience to both user account & admin account.
  • Integrate data modeling and machine learning to generate predictive diagnostic reports and ensure a seamless user experience.
  • Implement a quality review process (Greenlight Guru service) to ensure compliance with FDA requirements and ISO standards.
  • Conduct UAT to make sure all functionalities are working as expected. Gather feedback from users to refine the product.

My Responsibilities

Sole Designer, partially perform as product manager.
  • Stakeholder interviews
  • Product goals definition
  • Scope planning
  • User research & testing
  • Product design
  • Interactive prototyping
research

Clarify Ambiguities

I wanted to better understand

  • Who are the key users, and what do users need to achieve with the product?
  • How do users currently perform tasks related to lung cancer diagnostics?
  • Are there specific standards or regulations that the product must adhere to?

Research Insights

Stakeholder Interview Analysis

1

Algorithm Depends on Accurate Medical History

The algorithm's reliability and diagnostic capability heavily depend on accurate and comprehensive medical history data.

2

Clarify Functional Boundaries

The system can offer functionality to assist with recommendations (e.g., swab test report, scheduling or frequency suggestions), but it must not make autonomous clinical decisions.

3

User Need for Platform Manageability

Uncovered that stakeholders need admin control panel for platform manageability.

User Interview Analysis

Complex Process Complaint

The algorithm's reliability and diagnostic capability heavily depend on accurate and comprehensive medical history data.

Clarify Functional Boundaries

The system can offer functionality to assist with recommendations (e.g., swab test report, scheduling or frequency suggestions), but it must not make autonomous clinical decisions.

User Need for Platform Manageability

Uncovered that stakeholders need admin control panel for platform manageability.

User Need for Platform Manageability

Uncovered that stakeholders need admin control panel for platform manageability.

Compliance Experts Interview Analysis

  • Traceability and Audit Trails
  • Data Transparency
  • User Education
Define

Who need early cancer detection and how?

The goal of the research was to identify and understand customer needs with S1 Platform. SMD stakeholders have gain valuable user data to address scenarios in adopting S1 platform to achieve their goals. I initiated interviews with stakeholders to exam each user role's pain points, elicit their motivations for using the platform, and understand their needs. 

Main People Problems

1

Inefficient Workflows

The intuitive and user-friendly design of Elements_Efi has been specifically crafted to reduce the learning curve for users, making it accessible to people of all skill levels and backgrounds.

2

Trust in AI and Automation

Users concern about the reliability, transparency, and interpretability of AI-generated results, leading to hesitation in adoption.

3

Regulatory and Compliance Concerns

Requestors and Admins struggle with ensuring the platform adheres to strict medical and data regulations, creating additional workload and anxiety.

MVP Features

After multiple rounds of in-depth interviews, I identified and distilled the key business requirements into actionable product features.

1
Swab Test Creation
2
Participants Registration
3
Medical History Forms
4
Laboratory Process
5
SLF Generator
6
Admin Management

Visualising the end‐to‐end and finding opportunities for improvement

Gaining a complete picture of a customer's experience before, during, and after purchasing a swab test, I identified several pain points that need to be addressed after MVP.

design principle

Next Generation Cancer Diagnostics

Having fully understand the product, I conceptualized the following principles to guide design decisions.

1

Testing seamlessly

The overall workflow of a swab test should be seamless. Three different users are involved in the overall test flow, and there are both online and offline steps. We need to ensure there is no gap in between, and that data flow is uninterrupted.

2

Compliance

Follow the target market (European) regulations, ensuring compliance during design process of a medical device is vital so that the product can release to the market without barrier.

3

Start simple, scale up

We should thoughtfully restrict a user’s initial options as a gentle introduction to the product. As they go through the experience and gain familiarity, we should reveal more advanced filtering and customization options to them.

Framework

Setting the Design Direction

Using a basic visual layout, I attempted to convey a sense of how the platform is laid out and how the key features are implemented. This serves as the basis for the design process. I have created three low-fi workflows, the image below is one of them.

CONCEPTUALIZATION

Visualizing the swab test process

Laboratory Process
Our goal is to get client on board after testing MVP. Giving the complexed laboratory process, I felt the best way to accomplish this was to visualizing the workflow to make the process as simple and pain-free as possible. Also, any hint messages that could help user faster the process and reduce ambiguous is also suggested because we want the user felt happy during their exploration.
Swab Sample Holder
In correspondence with the off-line process, we mirrored the swab sample holder and created a virtual one. It was programmed to simulate the same process as the off-line process. By clicking on the sample hole, the system will auto-generate a UID for the swab test, and set related data. After set up, user can click Generate SLF button to generate the slf. file for cancer detection analysis.
Generate IR Cell Marker Positions File
In the third step of the processing workflow, the SLF generator is used to generate IR cell marker positions files. In order to generate the correct file, we iterated several times to make sure the right data was being filed.
Test Reports
By transforming complex data into clear, actionable insights, I ensure that design decisions are well-informed and user-centric.
Version 1.0

Polishing and adding functionality

UAT Feedback 1
Completing medical history form is a vital step of swab test creation process, but it is time consuming.
The stakeholders were able to connect one usability testing partner, a very experienced clinical professional (User 1) before MVP launch. We worked closely with her to help define tasks, establish objectives and evaluate the app.
During the testing, the user revealed workflows that does not fit the real working process, and usability issues. Physicians had to follow steps containing long medical related questionnaires in order to create a new swab test for a participant, which is time consuming. To be more specific, the initial requirement addressed the importance of filling the medical history forms before getting a swab test. However, user 1 pointed out that the medical history form is time wasting and a 45-minute test creation process is not practical in real situation. Plus, some of the questions cannot be retrieved immediately, which hinders physicians' ability to perform the swab test effectively.
Below is the original test creation workflow.
Solution: The medical history filling is no longer a mandatory step, users can complete it later but before shipping.
After few rounds of brainstorming with user 1 and stakeholder, we found that it is possible to have user save the medical history form filling in a later time, as long as the lab has not received the swab. In this case, physicians could complete the test creation process within 5 minutes, and there will be no medical risks if the process hold the swab from sending out before getting a complete medical history form.
UAT Feedback 2
Missing multiple short test operation forms in the new swab test process
User 1 revealed that, participants need to fill two extra forms after taken the swab test, participant test procedure form and post-test acceptance form. However, the original test creation process do not contain multiple steps, all user need to do in S1 system is filling out the medical history form. This change need to be reflected in the system in an efficient way before the release day.
Solution: Refining workflow for creating a new swab test  
After discuss with stakeholder, creating new test is a frequent option for requestors (such as doctors), I suggest to set this design change as a top priority and redesign the overall new test request workflow.
Using the workflow stepper, users are provided with an expectation of where they are in the process and what steps remain.
outcome

2.2x increase in the user engagement

After implementing the refined workflow for Sierra Medical Platform, user engagement increased by 2.2x. By addressing the challenge of lengthy and impractical medical history form completion, we reimagined the process to allow users to save the form as a draft and complete it later, provided it was finalized before swab shipping. This change reduced the swab test creation time from 45 minutes to just 15 minutes, aligning the workflow with real-world practices. The streamlined approach significantly enhanced usability, enabling physicians to focus on test accuracy and efficiency, while maintaining compliance with medical requirements. This user-centered adjustment not only improved the platform’s overall efficiency but also strengthened user trust and satisfaction.

3X

Registration Efficiency

2.2x

User Engagement

100%

Early Adopter Onboarding

reflection

Building the human-centered AI tool

Understanding Real-World Challenges

One key insight was the importance of designing for real-world workflows rather than theoretical assumptions. During user acceptance testing (UAT), involving clinical professionals revealed critical gaps in the initial workflows. The process of completing medical history forms, while essential, was unnecessarily time-consuming and misaligned with the operational constraints of physicians. By deeply engaging with users, we identified that allowing flexibility in form completion, as long as critical steps were verified before shipping the swab, could streamline the process without compromising medical accuracy.

The Best Way is the Simplest

Simplifying the complex was a cornerstone of this project. By introducing the option to save medical history forms as drafts, we enabled physicians to focus on critical steps first and revisit secondary tasks later. This streamlined the test creation process from 45 minutes to just 15 minutes. This approach not only addressed the challenge of time constraints but also showcased how small, intuitive changes can deliver massive improvements in user experience. The simpler workflow reduced friction, minimized errors, and empowered users to work more effectively.

Collaboration as the Key to Success

Collaboration with stakeholders, including clinicians like Dr. Forman, played a pivotal role in ensuring that every design decision was validated against practical constraints and user goals. This iterative feedback loop not only refined the platform's functionality but also fostered trust among users.