In## 💡 Inspiration
We wanted to build an accessible, non-invasive health screening tool that allows users to monitor critical biometric trends early, right from their own devices. Many serious conditions show subtle early warning signs—like a yellowish hue in the eyes or variations in sleeping heart rate—that patients often miss. VitalsPulse AI was born to bridge this gap, acting as a smart, preliminary check-up assistant.
⚙️ What it does
VitalsPulse AI is an interactive digital health assistant that merges proactive and passive monitoring:
- Visual Analytics (Selfie Scan): Users can upload images or short videos to scan for visual indicators of conditions like microscopic jaundice, conjunctivitis (such as Madras Eye), anemia paleness, or orbital swelling (edema).
- Vocal Biomarker Analysis: A dedicated space designed to evaluate vocal patterns for respiratory or vocal cord distress.
- Wearable Sync: A background portal that simulates pulling passive data (like Heart Rate Variability and sleep metrics) from platforms like Apple HealthKit to track continuous trends.
- Triage Dashboard: Aggregates these data streams to give users a clear, visual overview of their health.
🛠️ How we built it
The web application is built entirely using Python and Streamlit for a responsive, modern frontend UI. We integrated Plotly and NumPy to power the telemetry data visualizations, ensuring the dashboard looks clean and clinical. To make our working prototype dynamic for judges, we built an smart keyword-matching engine that analyzes uploaded file names (such as identifying conjunctivitis symptoms when a "Madras Eye" image is uploaded) and shifts the UI warning states automatically.
🚧 Challenges we faced
We hit quite a few roadblocks, starting with local environment issues! Getting Python 3.14 to run Streamlit properly on our Windows path required some troubleshooting. Once the server was live, we encountered file handling issues—specifically a crash where our image decoder tried to parse an uploaded .mp4 video file. We overcame this by rewriting our media logic to cleanly distinguish between images and videos dynamically.
🎉 Accomplishments that we're proud of
We are incredibly proud to have built and deployed a fully functional, cloud-hosted MVP (Minimum Viable Product) from scratch over a single weekend. Navigating from a completely empty terminal to a live, working URL on Streamlit Community Cloud was a massive achievement for us!
🔮 What's next for VitalsPulse AI
Our next step is moving past the prototype stage. We plan to connect the image uploader to a real convolutional neural network (CNN) model for clinical-grade image processing, integrate the Wearable Sync tab with the actual Apple HealthKit and Fitbit APIs, and implement strict end-to-end data encryption to meet healthcare standards.
Log in or sign up for Devpost to join the conversation.