NutriAI

"NutriAI : Redefining how we see, eat, and live. Powered by GEMINI "

Comment 
# About NutriAI: Precision Nutrition at Scale

## The Inspiration :
The journey of NutriAI began with a simple observation: while we live in an era of hyper-convenience, maintaining a healthy diet remains a complex, manual chore. Most people give up on nutrition tracking not because they lack discipline, but because the friction of logging every ingredient is too high. 

Inspired by the emergence of **Multimodal AI**, we envisioned a world where a single photo could replace a dozen search queries. We wanted to build more than a tracker; we wanted to build a "Nutritionist in your pocket" that understands not just the food on your plate, but the person behind the screen.

## How I Built It :
NutriAI is built on a modern, decoupled architecture designed for speed and premium user experience:

- **The Brain (AI Integration)**: We utilized **Google Gemini 2.5-flash** for its industry-leading multimodal capabilities. By calling the API directly via HTTP, we achieved a low-latency pipeline that can identify complex meals, estimate serving sizes, and provide real-time coaching.
- **The Engine (Backend)**: A Node.js and Express environment handles the heavy lifting, managing secure image processing via `multer` and orchestrating the flow between the user's profile and the AI's logic.
- **The Interface (Frontend)**: To achieve a "state-of-the-art" feel, we avoided generic frameworks and built a custom UI using **Vanilla JavaScript** and **Modern CSS**. We implemented a sidebar-based dashboard with split-layout panels to create a cinematic, high-end aesthetic.

## The Science (Math & Formulas) : 
To provide accurate, personalized advice, NutriAI implements standard nutritional science formulas using LaTeX-supported logic:

### Body Mass Index (BMI):-
We calculate the user's BMI to categorize their physical profile and adjust calorie targets:
$$BMI = \frac{weight (kg)}{height (m)^2}$$

### Basal Metabolic Rate (BMR):-
The daily calorie target is derived from the **Mifflin-St Jeor Equation**, ensuring that recommendations are grounded in human biology:
$$BMR = (10 \times weight_{kg}) + (6.25 \times height_{cm}) - (5 \times age) + 5$$

## Challenges Faced :
- **Multimodal Synchronization**: The biggest challenge was ensuring the AI understood the *context* of the user's profile (like muscle gain goals) while analyzing a *static image* of a salad. We solved this by implementing a "Context-Aware Prompting" system that injects the user's life-stats into every AI request.
- **UI Performance**: Creating a premium design with smooth animations and complex background panels without slowing down the app required deep optimization of CSS paint cycles and asset loading.
- **Data Persistence**: Synchronizing local storage state with real-time AI responses to maintain a consistent "Food Log" required a robust state management system built from scratch in Vanilla JS.

##  What I Learned :
This project was a deep dive into the **Gemini Ecosystem**. I learned how to move beyond simple text prompts and leverage Multimodal Large Language Models (MLLMs) for real-world utility. I also refined my skills in **"Product-First" Engineering**, realizing that high-end aesthetics are just as important as the code for user trust in a health-focused application.

---
**NutriAI** — *Redefining how we see, eat, and live.*

Built With

Share this project:

Updates