The Intelligent Dietary Companion
An interactive blueprint exploring the architecture, data flows, and AI models of the NutriGenius AI personalized dietary system.
🏛️System Architecture
NutriGenius AI uses a microservices architecture for scalability and maintainability. Hover over a service in the diagram below to learn more about its specific role and technology.
Hover a Service
Detailed information about the selected service will appear here.
🗺️The Data Journey
Follow the path of data from user onboarding to receiving a personalized meal plan.
Onboarding & Data Collection
The user registers and provides profile data (biometrics, goals, medical history) via a guided wizard. Data quality is ensured with client and server-side validation.
Secure Storage
The validated profile is sent via secure HTTPS to the User Profile Service. Derived metrics like BMI are calculated and the profile is stored in the MongoDB database.
Recommendation Request
The API Gateway authenticates the user via JWT and forwards the request to the ML Inference Service.
AI Generation
The ML service fetches the user's profile and interaction history, feeding it into the hybrid model to generate a ranked list of suitable recipe IDs.
Displaying Results
Recipe IDs are "hydrated" with full details (name, image, nutrition). This rich data is sent to the frontend and rendered in an interactive dashboard.
🧠The AI Core
The system's intelligence is built on a foundation of high-quality data and a hybrid recommendation strategy. Explore the components below.
Data Corpus Sources
The system integrates multiple datasets to build a comprehensive and culturally diverse recipe knowledge base.
Hybrid Recommendation Model
The engine blends two methods to provide recommendations that are both relevant and novel, overcoming the "cold start" problem.
⚙️Algorithms
A look under the hood at the core algorithms that power our AI and the strategies we employ to overcome common development hurdles.
Content-Based Filtering
Recommends recipes based on their inherent properties. We use TF-IDF Vectorization to convert recipe text (ingredients, cuisine) into numerical vectors and Cosine Similarity to find items similar to a user's preferences.
Collaborative Filtering
Suggests recipes based on the behavior of similar users. We employ Matrix Factorization (SVD) to uncover latent taste profiles from the user-recipe interaction data (ratings, saves).
Nutrient Deficiency Forecasting
Predicts nutrient risk levels. A Random Forest Classifier is trained on public health data (NHANES) to classify a user's risk as low, medium, or high based on their dietary intake patterns.
Cooking Time Prediction
Estimates prep and cook times from recipe text. We fine-tune a BERT (Transformer) model, a powerful NLP technique that understands the context and semantics of cooking instructions.
✨Advanced Features
Cutting-edge integrations transform the application from a simple tool into a dynamic and intelligent cooking companion.
LLM-Powered Recipe Generation
The system uses Generative AI to create full, personalized, step-by-step recipes on-demand, injecting safety constraints for allergies and health conditions directly into the prompt.
Hyper-Local Recommendations
By using the user's location, the system boosts recommendations for traditional and seasonal dishes from their specific region, creating a culturally resonant and delightful experience.
👥Meet the Team
The dedicated individuals behind NutriGenius AI.
Vedant Bhor
Project Lead
Tanay Hingane
AI/ML Specialist
Adarsh Tile
Backend Lead
Yadnesh Udar
UI/UX Designer
📚Citations
This work is inspired by and builds upon recent advancements in personalized nutrition and intelligent health systems.