💡 Inspiration The idea for Dermo-Gemini Pro was born from a common experience we’ve all shared: "Cyberchondria."
We noticed that whenever someone finds a strange rash or skin lesion, they immediately turn to Google Images. This almost always leads to panic, misinformation, and unnecessary anxiety. On the other hand, booking an appointment with a dermatologist can take weeks.
We asked ourselves: Is there a middle ground? Could we build an intelligent, accessible tool that bridges the gap between "Googling it" and "Seeing a Doctor"? We wanted to create a calm, evidence-based triage assistant that uses the power of AI to provide immediate guidance and peace of mind.
💻 What it does Dermo-Gemini Pro is a multimodal medical triage application. It allows users to:
Securely Login to a simulated doctor/patient portal.
Upload a photo of a skin lesion (Visual Input).
Describe their symptoms (Textual Input).
The app processes both inputs simultaneously to generate a structured medical assessment, offering visual observations, potential conditions, and a clear Urgency Rating (Low, Moderate, High).
⚙️ How we built it This was our first time competing in a hackathon, so we adopted a "90% Gemini" architecture. Instead of writing complex computer vision algorithms from scratch, we leaned entirely on the Google Gemini API.
The Brain: We used gemini-flash-latest because of its speed and multimodal capabilities. We engineered a specific "System Prompt" that forces the model to act as a Dermatologist Assistant, strictly adhering to safety protocols.
The Body: We built the frontend using Streamlit (Python). This allowed us to create a responsive, interactive dashboard without needing HTML/CSS expertise.
The Polish: To make the UI feel professional, we implemented Session State for user authentication and integrated Lottie Animations to create a welcoming, modern aesthetic.
🚧 Challenges we faced As beginners, the road was not smooth! We faced several significant technical hurdles:
The "Black Screen of Death": Early on, our Streamlit app refused to load in the browser. We spent hours debugging WebSocket conflicts and port issues before realizing we needed to adjust our server configuration.
Version Conflicts: We initially tried to run the project on the experimental Python 3.14, which caused major compatibility issues with the protobuf library. Downgrading to a stable Python 3.12 environment was a crucial lesson in dependency management.
Model Availability: We struggled to find the correct model name that supported image generation for our API key, eventually discovering that gemini-flash-latest was the most reliable solution for our region.
🧠 What we learned This project was a crash course in Full-Stack AI Development.
We learned how to integrate Multimodal AI, processing text and images in a single API call.
We mastered Prompt Engineering, realizing that how you ask the AI is just as important as the code you write.
Most importantly, we learned that coding isn't just about syntax—it's about debugging, perseverance, and solving real-world problems.
🚀 What's next for Dermo-Gemini Pro In the future, we plan to:
Integrate a real backend database (like Firebase or MongoDB) to save patient history.
Add a "Find a Doctor" feature that uses Google Maps API to locate the nearest dermatologist if the urgency rating is "High."
Log in or sign up for Devpost to join the conversation.