NutriSim

AI nutrition counseling simulator for 8,100+ Japanese dietitian graduates. Built by a practicing dietitian using Bolt.new. AI patient conversations with real-time skill evaluation and feedback.

Comment

🧠 Inspiration

I am a registered dietitian, not a programmer.

My journey into tech started from a single, critical observation in my clinical practice:
👉 Our best and brightest new dietitians are graduating academically brilliant, but practically terrified.

In Japan, while 8,100 new dietitians graduate annually, the curriculum offers a stark contrast to global standards.
🇺🇸 US dietitians undergo over 1,000 hours of supervised practice;
🇯🇵 Japanese students receive only a fraction of that.

This isn't just a gap — it's a chasm between knowledge and confidence.

This challenge resonated deeply with my personal experience:

When I first started my career, I struggled with the same issue —
Despite having solid nutritional knowledge, I felt anxious and unprepared when facing real patients with complex emotional and cultural barriers.

Through mentoring new graduates, I consistently observed this same pattern repeating.

🎓 Many expressed confidence in their nutritional knowledge,
but uncertainty in how to:

  • transform that knowledge into practical communication
  • deliver the right message at the right time to the right patient

The real challenge isn't just knowing the science.
It’s the art of translating science into human conversation —
adapting to confusion, respecting culture, and delivering advice with empathy — all in real time.

🧾 This is a skill that cannot be learned from a textbook.

💡 Having experienced this firsthand and seen it in the next generation of dietitians,
I realized the urgent need for a safe, realistic practice environment
where essential counseling skills could be developed before meeting real patients.

🛠️ What it does

NutriSim is an AI-powered patient simulation platform that provides safe, realistic practice environments for nutrition counseling skills.

🔑 Core Functionality

  • 🗣️ Real-time voice conversations with AI patients using Web Speech API
  • 🧍‍♂️ 3 distinct patient cases:
    • Diabetes: Mr. Yamada (65)
    • Hypertension: Ms. Suzuki (78)
    • Metabolic Syndrome: Mr. Tanaka (42)
  • 🎯 5-axis evaluation system:
    • Listening skills
    • Professional knowledge
    • Empathy
    • Communication
    • Practical planning
  • 📊 Comprehensive learning analytics:
    • Radar charts
    • Detailed feedback
  • 📈 Adaptive difficulty based on user performance

👥 Realistic Patient Scenarios

Each AI patient has:

  • Rich medical histories and personality traits
  • Emotional states like resistance, confusion, acceptance, motivation
  • Cultural context based on Japanese dietary habits and social norms
  • Realistic challenges like medication adherence, family dynamics, and lifestyle barriers

⚙️ How we built it

🧩 Primary Platform: Bolt.new

As a dietitian with zero formal coding experience, Bolt.new wasn't just a tool —
it was my co-founder.

It empowered me to transform 7 years of clinical insight into a functional, production-ready application — in just 5 days.

🧱 Technical Architecture

  • Frontend: React 18 + TypeScript + Tailwind CSS
  • AI Engine: Gemini 2.5 Flash API
  • Voice: Web Speech API + Speech Synthesis API
  • Database: Supabase
  • Deployment: Netlify (CI/CD with Deploy Challenge)

🧪 Development Process

  • Persona Design: Based on 7+ years of clinical experience
  • Conversation Flow: Realistic nutrition counseling mapped into dialog
  • Prompt Engineering: For natural, emotionally intelligent responses
  • Evaluation Logic: 5-axis feedback system
  • UI/UX: Designed for ease, clarity, and minimal cognitive load

🚧 Challenges we ran into

🎭 AI Response Authenticity

Creating realistic patient behavior required balancing:

  • Emotional depth (frustration, resistance, motivation)
  • Cultural nuances (Japanese cuisine and etiquette)
  • Medical correctness (treatment reactions)

Solution:
Multi-layered prompt engineering with psychology-backed response models.

🎤 Voice Recognition Accuracy

Nutrition terms like “glycemic index” and “antioxidants” weren’t always captured.

Solution:
Confidence threshold settings + text input fallback for seamless usability.

🧠 Real-Time Evaluation Logic

Building immediate feedback that’s meaningful involved:

  • Conversation analysis
  • Knowledge validation
  • Empathy and clarity assessment

Solution:
A hybrid system using AI-powered interpretation + rule-based logic.

🏆 Accomplishments that I'm proud of

🧑‍💻 Technical Achievements

  • 95% feature completeness in just 5 days
  • Zero-downtime deployment via Netlify CI/CD
  • Fully responsive: desktop, tablet, and mobile
  • Scalable for concurrent users

📘 Educational Impact

  • Quantifiable skill growth with 5-axis feedback
  • Analytics for self-reflection and development
  • No technical knowledge required to use

🧠 Innovation Recognition

  • World’s first fully functional AI simulator for dietitian counseling
  • Combines:
    • Voice AI
    • Behavioral education
    • Clinical realism

🎓 What I learned

🔓 The Democratization of Creation

With tools like Bolt.new, domain experts can be builders — even without engineering backgrounds.

💡 Technical Insights

  • Modern AI (like Gemini) can simulate human nuance
  • Good UX comes from clinical empathy, not just interface design
  • Conversational design = code + psychology

🎯 Domain Knowledge as an Asset

My experience as a nutritionist made me uniquely suited to:

  • Build realistic personas
  • Define useful evaluation metrics
  • Understand the learner’s mindset

⚡ Rapid Prototyping

User testing feedback from fellow dietitians enabled real-time iteration —
something traditional dev cycles can’t match.

🚀 What's next for NutriSim

📈 Market Expansion

  • Target: 20,000–40,000 users across Japan
  • Pricing: ¥480/month (individual); institutional packages available
  • Growth: High potential in academic and healthcare sectors

🗺️ Feature Roadmap

  • New cases: renal, pediatric, and sports nutrition
  • Multilingual support for Asia
  • VR: immersive counseling simulation
  • Institutional dashboards for curriculum optimization

🤝 Partnership Strategy

  • 140+ nutrition programs nationwide
  • Hospitals for in-service dietitian training
  • Japan Dietetic Association and allied organizations

Built With

  • gemini-2.5-flash-api
  • lucide
  • netlify
  • postgresql
  • react-18
  • recharts
  • speech-synthesis-api
  • supabase
  • tailwind-css
  • typescript
  • vite
  • web-speech-api
Share this project:

Updates