Kinexis

Logo
Homepage
Instructions Screen
Camera Set Up Interface
Webcam Tracking 1
Webcam Tracking 2
Final Report
Meet the Team (Our Mission Screen)

Inspiration

We wanted to make physical therapy smarter, faster, and more accessible. Clinics spend hours manually measuring patient progress, and patients rarely see objective proof of their improvement. We imagined a solution where a simple webcam could replace expensive motion-tracking systems, turning any laptop into a professional-grade PT assistant.

In speaking directly with physical therapists, we heard firsthand how time-consuming and tedious it is to capture accurate measurements. Many described the process as a bottleneck, slowing down care, limiting patient engagement, and adding administrative burden. Standard daily treatment notes typically consume 15–20 minutes per patient visit, and comprehensive progress reports can take 30–45 minutes per patient, depending on complexity and payer requirements. In fact, documentation and administrative tasks can account for up to 49% of a provider’s workday, rivaling the time spent in direct patient care (SpryPT). These numbers highlight just how much time is lost to manual tracking, time that could be reclaimed for hands-on care and better outcomes. Our solution aims to automate and streamline this process, giving clinicians real-time insights and patients clear proof of their progress.

What does it do

Kinexis is a guided physical therapy platform that blends clinical precision with a patient-friendly experience. Using webcam-based skeletal tracking, it measures joint angles, repetitions, and range of motion during simple rehab exercises. A friendly on-screen mascot leads each movement, helping reduce anxiety and making the process feel supportive rather than clinical. This humanized interface is especially valuable in medical settings, where patients may feel intimidated by unfamiliar technology. By softening the experience without sacrificing accuracy, Kinexis promotes better engagement and compliance.

The system captures objective data in real time, analyzing only skeletal form to preserve autonomy and minimize unnecessary data collection. After each session, Kinexis generates a personalized medical report that includes the patient’s name, performance metrics, and recommended future exercises based on their results. Traditionally, physical therapists spend around 20 minutes manually measuring and documenting this data, meanwhile, Kinexis completes the same process in just one minute, cutting time usage by 95%. Designed to fit seamlessly into clinical workflows, it saves time while enhancing clarity and consistency in patient care. Kinexis is efficient, accurate, and empathetic, built to support both providers and patients.

How we built it

Kinexis was built as a fully client-side web application, designed for speed, privacy, and clinical-grade accuracy. The frontend is powered by HTML, CSS, and JavaScript, deployed via Vercel for fast, serverless delivery. We used Flask with Flask-SocketIO and Gunicorn during development to simulate real-time interactions and manage local testing environments. Eventlet handled asynchronous communication, while Flask-SQLAlchemy provided ORM flexibility across SQLite and PostgreSQL databases. For skeletal tracking, we integrated MediaPipe with OpenCV, using NumPy and Pandas for data processing and Matplotlib for visualizations. The system captures joint angles and repetitions in real time, all processed locally in the browser to preserve user privacy. We used jsPDF and ReportLab to generate downloadable PDF reports, complete with body diagrams, progress charts, and insurance-ready language.

The simulation runs autonomously with smooth animations and dynamic feedback. We used Python for backend logic and python-dotenv for secure environment variable management. Testing was handled with Pytest, and version control was managed through GitHub. The entire experience is designed to run without logins, backend dependencies, or cloud integrations, making Kinexis lightweight, portable, and fast.

Challenges we ran into

One unexpected hurdle was a version mismatch between our development environment and one of the frameworks we relied on. We were building in Python 3.14, but MegaPipe, a key framework in our pipeline, only supported Python 3.12. That meant we had to roll back our environment, debug compatibility issues, and rework parts of the model to function under the older version. It was a frustrating detour, but it taught us a lot about dependency management and keeping our stack flexible.

We also ran into a tricky merge conflict during a morning push. Two team members had made overlapping changes to the skeletal tracking module, and resolving the differences required a deep dive into both versions to preserve accuracy and performance. It slowed us down temporarily, but it reinforced the importance of tighter coordination and clearer commit messages when working on critical components.

Accomplishments that we're proud of

One of our proudest achievements is the visual experience. Every graphic and animation — from the mascot’s gestures to the interface transitions — was handcrafted to strike the right balance between warmth and professionalism. It’s friendly enough to put patients at ease, yet polished enough to feel credible in a clinical setting. We also created the sound design ourselves, using subtle audio cues to reinforce feedback and guide users through each step.

We’re proud of how smoothly the webcam integration worked. Kinexis runs entirely client-side, yet it was able to generate a real-time skeletal model that matched each user’s proportions and measured joint angles with solid accuracy. It tracked reps and range of motion reliably on standard hardware, all through a browser with no setup, making the experience feel simple yet responsive.

What's next for Kinexis

Kinexis was built to be fast, autonomous, and clinically useful — no backend, no logins, and no integrations. Moving forward, we plan to expand the exercise library to support more joints and movement types, allowing providers to assess a broader range of conditions with the same streamlined experience. We’re also exploring adaptive feedback, where the system recommends future exercises based on performance metrics captured during each session.

One of our key goals is to make Kinexis usable not just by clinicians, but also by insurance providers. By expanding the exercise library, refining report formats, and aligning outputs with payer expectations, we aim to support documentation that validates medical necessity and streamlines reimbursement, all while keeping Kinexis lightweight, private, and portable. Future updates will focus on customizable reports, multi-exercise sessions, and smarter feedback, ensuring Kinexis remains a fast, empathetic tool that delivers actionable insights in under a minute.