DermAid
Intelligent Acne Treatment System
Created on 16th October 2025
•
DermAid
Intelligent Acne Treatment System
Description of your solution
DermAid - Intelligent Acne Treatment System
The Problem We're Solving
If you've ever dealt with acne, you know the frustration: what works for your friend doesn't work for you. You spend a lot of money trying different products, following generic advice from apps that recommend the same treatments to everyone. Meanwhile, if you have darker skin, most solutions aren't even trained to understand your skin tone properly, leaving you to deal with post-inflammatory hyperpigmentation that nobody seems to care about.
The truth is, acne treatment today is still stuck in the trial-and-error dark ages. We think that's unacceptable.

Our Solution
DermAid is the first truly intelligent acne treatment system that learns YOUR skin. Not someone else's. Yours.
Think of it as having a dermatologist who studies your skin every single day, remembers what worked and what didn't, and gets smarter about treating you with each passing week. But instead of costing ₹2000 per visit, it's right there in your pocket.
Here's what makes DermAid different:
Personalized Learning Engine: We don't give you generic advice. Our AI observes how your skin responds to different treatments and adapts your plan weekly. If something's not working, we switch it up. If something's working great, we double down on it.
Treatment Response Prediction: Before you waste money on a product, we predict whether it'll actually work for your specific skin type. We analyze patterns from thousands of similar skin profiles and tell you upfront: "This will reduce your acne by 60% in 4 weeks" or "This popular product will likely irritate you - try this instead."
Skin Tone Intelligence: We built specialized models for darker skin tones because we're tired of AI that only works for light skin. Our system understands post-inflammatory hyperpigmentation, prevents scarring on melanated skin, and recommends products that actually work across all Fitzpatrick scales.
Week-by-Week Adaptation: In week 1, we start with a baseline. By week 3, we've learned your triggers. By week 6, we've found your perfect formula. And we keep optimizing from there.

How It Works
- Take a photo of your skin daily (you'd check the mirror anyway)
- Log what you're using - products, lifestyle factors, everything
- Watch DermAid learn - our AI analyzes your responses and adapts your treatment plan
- Get predictions - we tell you what will work BEFORE you buy it
- See real results - personalized treatment that actually understands your unique skin
The magic happens in the background. Our system is constantly comparing your progress, identifying patterns, and refining your treatment protocol. It's like having a clinical trial running just for you.
Why This Matters
500 million people worldwide struggle with acne. The global acne treatment market is worth $15 billion. Yet most solutions are still one-size-fits-all. We're changing that.
More importantly, we're addressing a real equity issue in dermatology. Most AI models are trained predominantly on light skin, leaving people with darker skin tones underserved.
DermAid was built from the ground up to work for everyone.
Technical Approach
We're building DermAid as a Progressive Web App with React/Next.js frontend and TensorFlow.js for client-side AI inference. This means acne analysis happens directly in your browser - your photos never leave your device unless you choose to sync them.
Our computer vision model uses a MobileNetV2 backbone fine-tuned for acne detection across diverse skin tones, with custom classification heads for acne type identification and skin tone adaptation. Firebase handles the backend - Cloud Firestore for real-time treatment tracking and progress syncing.
Our agentic AI learning engine autonomously aggregates treatment responses across users to power predictive algorithms, while proactively adapting individual user profiles weekly based on observed skin reactions, making intelligent decisions without requiring user intervention.


Tracks Applied (1)
