Danilo Caffaro

Posted on Feb 27

I Built a Dietbox Alternative with AI Body Composition Analysis — Here's How

#ai #health #webdev #nutrition

What if you could analyze a patient's body composition just from 2 photos?

That's the question I set out to answer when building NutriAssess — and the results surprised me.

As someone who's worked closely with nutritionists, I kept hearing the same frustration:

"I'm paying $30–80/month for tools like Dietbox, and half the features I don't even use."

So I built an alternative. And I used Gemini Vision AI to do something those tools can't.

🩺 The Problem: Nutritionists Are Overpaying

Platforms like Dietbox, Evofit, and NutriSoft are the industry standard in Brazil and growing globally. They do the job — but:

💸 $30–80/month per subscription
📋 Clunky interfaces built for desktop-first workflows
🚫 No AI-powered visual body composition analysis
🔒 Patient data locked in proprietary silos

For solo nutritionists, small clinics, or practitioners just starting out, this is a significant cost with limited flexibility.

🚀 Introducing NutriAssess

NutriAssess is a free-to-try, AI-powered nutritional assessment platform that gives nutritionists everything they need — without the bloated subscription.

🔗 Try it free → https://nutri-assess.vercel.app

💳 Get lifetime access for just $12 → https://caffaro.gumroad.com/l/nutri-assess

🤖 The Tech: Gemini Vision AI for Body Composition

Here's where it gets interesting.

Traditional tools require manual entry of skinfold measurements, circumferences, and other anthropometric data — collected by hand with calipers and tape measures.

NutriAssess adds a 360° AI Visual Assessment powered by Google's Gemini Vision AI. Here's how it works:

Patient uploads 2 photos (front view + side/back view)
Gemini Vision analyzes posture, body fat distribution, muscle definition, and symmetry
The AI generates a structured clinical report with:
- Estimated body fat percentage
- Fat distribution pattern (android/gynoid)
- Muscle mass visual assessment
- Postural observations
- Nutritional risk indicators
- Personalized dietary recommendations

This isn't a gimmick — it's a genuine clinical screening tool that saves 20–30 minutes per consultation.

// Simplified example of the AI analysis call
const response = await geminiVision.analyze({
  images: [frontPhoto, backPhoto],
  prompt: CLINICAL_ASSESSMENT_PROMPT,
  schema: BodyCompositionSchema
});

📊 Scientific Formulas Under the Hood

Beyond AI, NutriAssess implements validated scientific protocols:

Body Fat % — Jackson & Pollock (1978)

// 7-site skinfold formula (men)
Density = 1.112 - (0.00043499 × ΣSkinfolds) + (0.00000055 × ΣSkinfolds²) - (0.00028826 × Age)

// Siri equation (1956) — converts density to body fat %
BodyFat% = (495 / Density) - 450

Basal Metabolic Rate — Harris-Benedict (1919, revised 1984)

// Men:
BMR = 88.362 + (13.397 × weight_kg) + (4.799 × height_cm) - (5.677 × age)

// Women:
BMR = 447.593 + (9.247 × weight_kg) + (3.098 × height_cm) - (4.330 × age)

Total Energy Expenditure

TEE = BMR × Activity Factor
// Sedentary: 1.2 | Light: 1.375 | Moderate: 1.55 | Active: 1.725 | Very Active: 1.9

These are the gold-standard formulas used in clinical nutrition — the same ones Dietbox uses — but now accessible for free.

✨ Full Feature List

Feature	NutriAssess	Dietbox
AI Body Composition from Photos	✅	❌
Anthropometric Calculations	✅	✅
Anamnesis (Patient History)	✅	✅
Patient Portal	✅	✅
Dietary Assessment	✅	✅
Nutritional Plan Generator	✅	✅
Progress Tracking	✅	✅
Price	Free / $12 lifetime	$30–80/month

🔬 Core Modules

1. 360° AI Visual Assessment
Gemini Vision analyzes patient photos for body composition, posture, and fat distribution. Generates a complete clinical report in seconds.

2. Anthropometric Calculations
IMC, % body fat (Jackson & Pollock 3/7 sites), BMR (Harris-Benedict), TEE, ideal weight, lean mass, fat mass — all calculated automatically.

3. Anamnesis Module
Structured patient history: medical background, eating habits, physical activity, sleep, stress, digestive health, and dietary restrictions.

4. Patient Portal
Patients access their own assessments, progress charts, and nutritional recommendations through a dedicated login.

5. Dietary Assessment
Detailed food intake analysis with macros, micros, and meal timing recommendations.

6. Nutritional Planning
AI-assisted meal plan generation based on patient goals, restrictions, and metabolic data.

🏗️ Tech Stack

Frontend: Next.js 14 + TypeScript + Tailwind CSS
Backend: Next.js API Routes + Prisma ORM
Database: PostgreSQL (via Supabase)
AI: Google Gemini Vision API
Auth: NextAuth.js
Deployment: Vercel

📈 The Business Model

I wanted NutriAssess to be genuinely accessible:

Free tier: Core assessment tools, limited AI analyses
Lifetime deal: $12 one-time (no recurring fees, ever)

👉 Grab lifetime access for $12 → https://caffaro.gumroad.com/l/nutri-assess

For context, that's less than a single month of Dietbox's cheapest plan.

🎯 Who Is This For?

Solo nutritionists and dietitians
Nutrition students building their first client base
Small clinics looking to cut SaaS costs
Personal trainers who work with nutrition
Anyone curious about AI-powered health assessments

🔮 What's Next

[ ] Mobile app (React Native)
[ ] TACO/IBGE food database integration
[ ] PDF report generation
[ ] WhatsApp patient reminders
[ ] Multi-practitioner clinic management

Try It Yourself

The app is live and working right now:

🔗 https://nutri-assess.vercel.app

If you're a nutritionist, dietitian, or just curious about what AI can do in clinical settings, I'd love your feedback.

And if you find it valuable:

💳 Lifetime access for $12 → https://caffaro.gumroad.com/l/nutri-assess

Built with ❤️ using Next.js, Gemini Vision AI, and a genuine desire to make nutritional care more accessible.

Have questions about the tech? Drop a comment below — I read everything.

DEV Community