UX Research · Design · Full-stack development

AwareHome SmartBathroom Dashboard

A dashboard that lets occupational therapists easily view live and historical bathroom sensor data, organized by participants, homes, and studies.

overview

Role: UX Researcher & Systems Designer
Team: TODO(nathan): team size and who you worked with (e.g. project lead Jon Sanford)
Timeline: May 2024 – Aug 2025
Tools: React, Flask, InfluxDB, Home Assistant, Grafana, Raspberry Pi

the problem

The bathroom is central to independence, yet age-related changes make daily tasks harder. To study these challenges, our team embedded four sensors under the toilet seat to capture data from older adults (60+) with physical impairments. Occupational therapists needed a dashboard to track this data over time, viewing both live and historical information across participants, homes, and studies.

The sensor-equipped toilet seat and the context it is used in

How might we design a solution that enables occupational therapists to easily view, interpret, and act on pressure sensor data collected from participants?

Occupational therapists lacked a centralized tool to visualize seat sensor data, making it difficult to monitor participants, compare across homes and studies, or generate insights for care.

research

Effort 1: Literature review & conversations. I reviewed past AwareHome projects and spoke with project lead Jon Sanford to understand recurring challenges. A key issue identified was the difficulty of controlling and accessing historical sensor data. The goal was to ground the project in existing research and uncover therapist pain points around messy, inaccessible data.

Annotated excerpt from the SMARTBATHROOM 102 study — Jones, B., Sanford, J., Presti, P., et al. (2022). SMARTBATHROOM 102: Studying bathing transfer performance. Georgia Institute of Technology.

Effort 2: Defining the data structure. I designed a clear, flexible data model that organizes sensor data and supports navigation between levels of participant, home, study, and seat, ensuring scalability with essential metadata.

Sticky-note map of how participants, homes, studies and seats relate

Scoping. Through our research, we scoped the project by prioritizing data organization and core visualizations as immediate needs for therapists, while identifying additional features.

Scoping chart separating immediate needs from later features

Methods: Literature reviewStakeholder conversationsTherapist interviewsData modelingScoping
Participants: 8 occupational therapists across 3 testing cycles

Key insights

Data access

Controlling and accessing historical sensor data was a recurring challenge.
Therapist pain points centered on messy, inaccessible data.

Structure

Navigate between participant, home, study, and seat.
Keep essential metadata like start and end dates for traceability.

Priorities

Data organization and core visualizations are the immediate needs.
Additional features were identified for later phases.

design process

Sketch
TODO(nathan): add an early sketch of the dashboard (paper or FigJam).
Iteration 1 · House-centric
Iteration 2 · Participant-centric

House-centric view, where homes are the top-level navigation — **The view.** I shifted to a participant-first structure, giving therapists a clearer, more intuitive way to track each person's data while still allowing filtering by home or study.

Participant-centric view, where each participant is the top-level navigation — **The view.** I shifted to a participant-first structure, giving therapists a clearer, more intuitive way to track each person's data while still allowing filtering by home or study.

Vertical layout with information stacked above the graphs — **The dashboard.** A sidebar to show key navigation allows the graph to expand across the full width, which keeps data upfront while keeping secondary information accessible.

Horizontal layout with a sidebar and full-width graphs — **The dashboard.** A sidebar to show key navigation allows the graph to expand across the full width, which keeps data upfront while keeping secondary information accessible.

The view. A participant-centered view makes metadata easy to access. Originally designed around homes, the first view assumed therapists needed direct visibility into each home. However, it limited their ability to focus on individuals and their specific data.

Participant-centric view

The dashboard. A sidebar layout surfaces key graphs upfront. The first version stacked information vertically, forcing therapists to scroll before reaching the graphs they needed most, which buried critical insights under less relevant details.

Sidebar dashboard

Feature 1: Metadata management. Therapists can easily manage participant metadata, including study details, start and end dates, and other key attributes. This feature was designed to reduce the clutter and inconsistencies of raw sensor data, giving therapists a clean and reliable way to keep records up to date.

Feature 2: Data filtering. Therapists can use a calendar to filter by day and sort through individual seat sessions. Each instance is linked to an interactive Grafana-powered graph, making it easy to explore both current and historical data by zooming and clicking data points.

Calendar filtering and session graphs

Feature 3: Compare data. Therapists can compare data across participants, sessions, or homes to identify patterns and anomalies. This comparative view was designed to help therapists make informed decisions more quickly, without needing to manually align spreadsheets or charts.

Comparing sessions

development

System architecture. Seat sensors stream data into a local InfluxDB instance on the Raspberry Pi, which syncs to a secure InfluxDB Cloud bucket. From there, a Flask API organizes and exposes the data, Grafana renders it into interactive charts, and the React dashboard embeds those charts for therapists, ensuring both real-time and historical data are accessible.

System architecture from seat sensors to the React dashboard

Data flow. To support this pipeline, I designed a relational data model connecting Participants, Seats, Homes, Studies, and Sessions, with metadata like start/end dates for traceability. The diagrams illustrate how data flows through the system and how functions interact to keep relationships consistent and scalable.

Raw historical sensor data, blurred for privacy — Raw historical data (blurred for privacy)

The same data displayed on the dashboard — Display of data on the dashboard

outcome

occupational therapists interviewed: 8
testing cycles: 3
RESTful endpoints designed and built: 20+

Solution
We created a dashboard that lets occupational therapists easily view live and historical bathroom sensor data, organized by participants, homes, and studies.
Impact for users
Therapists gained a clearer, more efficient way to track challenges and tailor interventions for older adults.
Impact for team
Our team learned how to bridge research, design, and development to deliver a clinically useful tool.

reflection

Future work includes deploying seats in participant homes at scale and expanding the dashboard with additional features for therapists and engineers, such as richer note-taking, alternative visualizations, or sensor setup tools. Getting the system hosted on Georgia Tech's servers was a long and challenging process, but it ensures that anyone continuing this project will have a much easier time building on top of our work.

I'm proud of how we got the groundwork running and excited to see how this platform can continue to grow in both technical capability and real-world impact.