Overview

Genetic Research Platform used by laboratories to run, analyze, and report on genetic tests commissioned by clinics, hospitals, biotech companies, and research institutions.

Problem

- Onboarding time for a new lab: ~3 weeks
- Turnaround for test reports: avg. 7.5 days
- Churn (small-to-mid-size labs): 28% in 6 months
- Clinician satisfaction (report clarity): 5.8 / 10
- Internal hypothesis: We were not scaling with user complexity

Design Challenge

Key Design Objectives:

- Automate routine tasks (tagging, filtering, reporting)
- Support clinical users with non-technical backgrounds
- Reduce onboarding/setup time for labs
- Drive adoption of premium features like cohort analysis

Hypotheses

The AI-Driven Solution

We considered implementing an AI-assisted onboarding and workflow builder tailored for researchers.

The AI model:
- Analyzed the user's dataset type, past experiments, and research objective (collected at sign-up).
- Recommended pre-built pipelines (e.g., CRISPR variant detection, protein folding prediction).
- Used NLP to parse dataset names and suggest metadata normalization.

Offered in-product explanations with AI-generated descriptions of each workflow step.

AI Feature Definition and UX Role

We co-created a feature framework with PM and ML teams:

User Research

Interviewed 13 users (bioinformaticians, lab techs, clinicians)
Shadowed support team handling common issues
Heatmaps revealed high dwell time on workflow setup and metadata tagging forms
Reviewed failed onboarding sessions: 40% stalled on first dataset

Key Pain Points:

Manual metadata tagging = cognitive load + high error rates
“Guesswork” when selecting panels and workflows
Reports too technical for non-geneticist clinicians
No insights into trends or outliers in patient cohorts

Key Findings and Insights

Design

Workflow Redesign from Manual to AI-Augmented

Before

No dynamic suggestions — user must manually:
- Choose sample type
- Select a gene panel from a long, alphabetized list
- Enter parameters (e.g., genome build, quality filters)
- No guidance or automation
- Confusing terminology for non-bioinformaticians
- High cognitive load and decision paralysis

After AI Integration

Replaced form fields with an intelligent assistant interface
Integrated AI-powered recommendations based on:
- Sample content
- Dataset info
- Lab history (frequently used panels)

AI Panel Recommendation

Created a panel recommender module embedded in the test setup screen.

Designed:
- A side-panel widget summarizing suggestions (with panel names, scores, rationale)
- “Why This?” modal for explainability
- Interaction patterns for AI generation and user acceptance

Integrated lab history context and patient phenotype filters into logic

Report Reimagination

Completely rethought the reporting patterns.
- AI interprets VCF, FASTQ and BAM files into clinician-readable insights
- Adds “Likely Pathogenic” flags with link-outs to evidence
- Uses language like: “A BRCA1 mutation has been detected, commonly associated with increased breast cancer risk.”
- Visual gene diagrams + AI-generated summary section

Cohort Intelligence Dashboard

Designed new dashboard for cohort-wide insights:
- Mutation distribution by gene (bubble chart, bar chart)
- Filters: age group, tissue type, phenotype, test type
- Real-time alerts from AI detecting unusual patterns

Key Takeaways

Human-Centered AI Design

Measurable Business Impact

+17% increase in customer retention among clinics and labs
NPS rose from 42 to 71, with usability tests showing a 66% faster interpretation time
Sales teams began using redesigned features as a demo asset, accelerating enterprise deal conversations by showing real-world utility.

Strategic Role of Product Design

Led end-to-end redesign: discovery → design → delivery, collaborating with AI/ML engineers, genetic counselors, and sales teams

Finally:

This project wasn’t about making the UI prettier — it was about reshaping workflows with intelligence, empathy, and clarity. As Lead Product Designer, I connected technical potential with human needs, ensuring that:
Complex AI became intuitive and trustworthy
Results were measurable and aligned with business goals
The product evolved into a strategic differentiator in the competitive landscape