AIdeas: DermAI Hope - Bringing Dermatology to 2.5 Billion People Without a Doctor
DermAI Hope puts AI-powered skin screening in your pocket: free, private, offline-capable, in 10+ languages. Photograph the affected area, answer 5 questions, and receive a diagnosis with confidence score, treatment guidance, and urgent care alerts — in seconds, no account required.
App Category
Social Impact
Team Members: @tonnybright, @kobiampem, @elikem
Tags:
#aideas-2025 #social-impact #EMEAMy Vision
The idea began with a number that stopped me cold: 4.69 billion.
That is how many people globally are affected by skin diseases every single year, according to the Global Burden of Disease Study 2021. Skin conditions account for 41.9 million disability-adjusted life years, meaning decades of human life lost to disability, suffering, and despair. Yet for more than 2.5 billion people in underserved regions, rural Africa, Southeast Asia, refugee camps, remote Latin America, a dermatologist is not a phone call away. A dermatologist is a fantasy.
The cruelty is not just physical. Skin diseases are visible. They are worn on the face, the hands, the neck. They invite stares, rejection, discrimination. Children are kept home from school. Adults lose jobs. Families fracture under stigma. The WHO Executive Board Resolution EB156(24), passed in February 2025, explicitly identifies this mental health burden (depression, anxiety, social isolation) as a compounding crisis that follows the physical disease like a shadow.
DermAI Hope is my answer to that number. It is a mobile-first web application that acts as an AI-powered dermatologist in your pocket: free, private, fast, and available offline. A user photographs their skin condition, answers five guided questions about their symptoms, and within seconds receives an AI diagnosis with confidence scores, treatment guidance in their native language, and a severity flag that tells them when to seek urgent care immediately.
I built this because I believe that access to diagnosis should be a human right, not a geographic privilege. And with Amazon Bedrock's Nova multimodal models, AWS Lambda, and a carefully designed serverless architecture, I proved it could be done within the AWS Free Tier.
Why This Matters
The numbers do not lie, but they do not bleed either. Let me make them real.
Imagine a mother in rural Kenya. Her four-year-old has had an itchy rash spreading across his arms for two weeks. The nearest clinic is three hours away by bus. It costs money she does not have. And even if she arrives, the clinic's doctor is a general practitioner, not a dermatologist. There are fewer than 0.5 dermatologists per 100,000 people across sub-Saharan Africa.
She does not know if this is eczema (treatable with a cheap cream) or scabies (highly contagious, requires family-wide treatment) or early impetigo (needs antibiotics before it spreads to her other children). She waits. It gets worse. The child scratches until he bleeds. He cannot sleep. He cries.
Now imagine she had DermAI Hope on her phone. She takes a photo. She answers five questions. In ten seconds, in Swahili, she reads: "This appears to be atopic dermatitis (eczema). Confidence: 87%. Apply a fragrance-free moisturizer twice daily. A low-potency hydrocortisone cream is available at most pharmacies. Monitor for infection signs."
That is dignity. That is hope. That is ten seconds that change everything.
Why this matters, point by point:
- Access Crisis: The WHO confirms a severe global dermatologist shortage. DermAI Hope provides 24/7 expert-level screening with zero specialist required.
- Early Detection Saves Lives: Melanoma caught early has a 99% survival rate. Caught late, that drops below 30%. Our severity flagging could be the difference between life and death.
- Stigma Reduction: Private diagnosis on your own phone means no judgment, no stares, no whispers. Privacy restores dignity.
- Economic Liberation: Free diagnosis eliminates catastrophic health expenses for families already living on the edge.
- Healthcare System Relief: If primary care workers can handle 80% of skin conditions with AI guidance, specialists can focus on the complex cases that truly need them.
- WHO Alignment: This project directly answers the WHO's 2025 call for "innovative integrated service delivery models, including telemedicine platforms and training for digital assessments." We are not just building an app. We are answering a global mandate.
How I Built This
Architecture Overview
DermAI Hope is a fully serverless, mobile-first Progressive Web App built entirely on the AWS Free Tier, developed with Kiro as the primary agentic IDE.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
User (Mobile Browser / PWA)
| HTTPS
Amazon CloudFront (CDN + HTTPS)
|
AWS Amplify (Next.js Frontend)
| API calls
Amazon API Gateway
| invokes
AWS Lambda (4 functions)
+-- upload-handler --> Amazon S3 (encrypted, 24h auto-delete)
+-- diagnosis-engine --> Amazon Bedrock (Nova Pro, multimodal)
+-- history-tracker --> Amazon DynamoDB
+-- translator --> Amazon Translate
Supporting Services:
Amazon Cognito (anonymous auth, privacy-first)
Amazon Rekognition (image quality check)
AWS KMS (image encryption)Demo
Screenshots: home screen, condition library, diagnosis history, diagnosis results.
What I Learned
Technical Insights
1. Multimodal prompting is both art and science.
Getting Nova Pro to give medically sound responses required iterative prompt engineering over dozens of sessions. The breakthrough was structuring the system prompt as a "clinical screener protocol," giving the model a role, constraints, and an explicit uncertainty acknowledgment mandate. A model that says "I'm not sure, please see a doctor" is infinitely more valuable than one that guesses confidently.
Getting Nova Pro to give medically sound responses required iterative prompt engineering over dozens of sessions. The breakthrough was structuring the system prompt as a "clinical screener protocol," giving the model a role, constraints, and an explicit uncertainty acknowledgment mandate. A model that says "I'm not sure, please see a doctor" is infinitely more valuable than one that guesses confidently.
2. Serverless architecture truly changes what's possible at zero cost.
A globally-available, AI-powered medical application with encryption, multi-language support, and offline capability, running for under $10/month. That is a paradigm shift for social impact applications. The WHO can't fund a dermatologist in every village. But a Lambda function doesn't need a salary.
A globally-available, AI-powered medical application with encryption, multi-language support, and offline capability, running for under $10/month. That is a paradigm shift for social impact applications. The WHO can't fund a dermatologist in every village. But a Lambda function doesn't need a salary.
3. Kiro's spec-driven development is the future of building.
Kiro forced me to think specification-first: define every component, every API contract, every data shape before touching code. The result was dramatically fewer bugs at integration time. Kiro's agentic mode then wrote the boilerplate from those specs, letting me focus on the AI integration.
Kiro forced me to think specification-first: define every component, every API contract, every data shape before touching code. The result was dramatically fewer bugs at integration time. Kiro's agentic mode then wrote the boilerplate from those specs, letting me focus on the AI integration.
4. Privacy-first is not a constraint. It is the product.
For stigmatized conditions, making account creation optional was not a technical decision. It was an ethical one. People will not share a photo of their skin condition if they fear it will be linked to their identity. By designing privacy in from day one, we removed the single biggest barrier to adoption.
For stigmatized conditions, making account creation optional was not a technical decision. It was an ethical one. People will not share a photo of their skin condition if they fear it will be linked to their identity. By designing privacy in from day one, we removed the single biggest barrier to adoption.
5. The 10-condition focus is a feature, not a limitation.
By focusing on the conditions that cause 80% of global burden, the model became dramatically more accurate. A focused tool that works is infinitely better than a comprehensive tool that doesn't.
By focusing on the conditions that cause 80% of global burden, the model became dramatically more accurate. A focused tool that works is infinitely better than a comprehensive tool that doesn't.
Human Insights
Building DermAI Hope was not just a technical journey. It was a reminder of why technology matters.
Reading through the WHO's EB156(24) resolution, pages of data about stigma, discrimination, mental health comorbidities, and children who never received care because their family didn't know what cream to buy, I felt something shift. This is not an app. This is a small piece of infrastructure for human dignity.
The 41.9 million disability-adjusted life years lost to skin disease are not statistics. They are people. The grandmother who stopped going to market because of her vitiligo. The teenager who dropped out of school because of severe acne. The infant whose scabies spread to her siblings because her mother didn't know it was contagious.
Every second of work I put into the diagnosis engine's accuracy was for them.
The best technology is invisible. When DermAI Hope works perfectly, the user doesn't think "wow, Amazon Bedrock is impressive." They think: I know what's wrong with my skin. I know what to do. I am not alone. That invisibility is the highest form of engineering.
DermAI Hope — because everyone deserves to be seen.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article