Return to Projects

Life Lens Suite

An AI-powered personal lifestyle companion that integrates food nutrition analysis, smart diet planning, and expense tracking—all from photo captures. Built with a Multi-Model AI architecture leveraging Google Gemini, Groq Llama 4, Cerebras, Deepgram, and Moondream for intelligent, real-time insights.

Visit Live Site Private Code

The Challenge

Modern lifestyles demand constant juggling between health consciousness and financial discipline. Individuals often struggle to maintain a balanced diet while keeping track of daily expenses. Manual calorie counting is tedious and error-prone, restaurant menus rarely provide nutritional guidance, and tracking scattered receipts for expense management is a logistical nightmare.

Life Lens Suite solves this by providing an intelligent, unified platform where a single photo can unlock comprehensive nutritional analysis, personalized diet recommendations, or instant expense categorization. By combining cutting-edge multimodal AI with voice recognition and OCR technology, users can effortlessly maintain their wellness and financial goals through natural interactions.

Core Features

🍽️

Food Lens - Instant Nutrition Analysis

Snap a photo of any meal and get instant calorie estimates, macronutrient breakdown (protein, carbs, fat), and ingredient identification. Results are validated against USDA FoodData Central for scientific accuracy.

📋

Menu Lens - Smart Restaurant Advisor

Photograph restaurant menus (supports multi-page) and receive AI-driven recommendations based on your budget and dietary goals. Features dual-mode analysis: Vision Mode (Groq Llama 4 Scout) for instant insights or OCR Mode (Moondream + Cerebras) for precision extraction.

🎤

Pantry Chef - Voice-to-Recipe

Record what ingredients you have via voice, and the AI generates personalized recipes. Powered by Deepgram for ultra-fast speech transcription and Cerebras for intelligent recipe reasoning.

📆

AI Diet Planner

Generate personalized 7-day meal plans based on your calculated BMR/TDEE (Mifflin-St Jeor equation), dietary preferences, and calorie targets. LLM-powered meal suggestions adapt to your lifestyle.

🧾

Finance Lens - Receipt Scanner

Photograph receipts to automatically extract merchant names, transaction dates, item lists, and totals. Track monthly spending with category breakdowns and historical summaries.

🤖

Multi-Model AI Orchestration

Intelligently routes requests to the optimal AI provider—Google Gemini for vision tasks, Groq for ultra-fast inference, Cerebras for complex reasoning, Deepgram for speech, and Moondream for edge OCR.

🔐

Secure Authentication

Google OAuth 2.0 integration with JWT-based session management. Features auto-refresh tokens, secure HttpOnly cookies, and token blacklisting for robust security.

⚙️

Enterprise-Grade Infrastructure

Built-in rate limiting (IP + Domain based), maintenance mode with graceful degradation, and health check endpoints. Fully containerized with Docker for consistent deployments.

Technology Stack

FastAPI High-performance Python web framework with automatic OpenAPI documentation, async/await support, and type validation via Pydantic.

Google Gemini (gemini-2.5-flash) Handles food image analysis, receipt OCR, and general text generation with native multimodal capabilities.

Groq (Llama 4 Scout) Processes menu images with blazing speed for real-time restaurant recommendations. 17B parameter multimodal model optimized for visual understanding.

Cerebras (GPT-OSS-120B) Powers complex diet plan generation and menu recommendation logic with 120B parameter model delivering sub-second responses.

Deepgram (Nova-3) Converts voice recordings to text for Pantry Chef feature. Supports Indonesian language with smart formatting.

Moondream Lightweight vision-language model for precise text extraction from menus and receipts without cloud dependencies.

LangChain Provides unified abstraction layer for switching between LLM providers with consistent interfaces and error handling.

PostgreSQL + AsyncPG ACID-compliant relational database with async connection pooling for high-concurrency operations. Supports Supabase for managed deployments.

Jinja2 Templates Server-side HTML templating with component-based architecture. Features a modern, responsive design with emerald-navy color scheme.

Docker + Docker Compose Complete containerized development and production environment with PostgreSQL service integration.

Vercel Production-ready deployment configuration for Python serverless functions with automatic scaling.

Yoyo Migrations Python-based database migration system integrated into application lifecycle for seamless schema evolution.

Key Advantages

Unified Lifestyle Management: One app covers nutrition tracking, diet planning, and expense management
Multi-Provider AI Resilience: Automatically falls back between 5 AI providers for maximum uptime
Real-Time Analysis: Sub-second response times powered by Groq and Cerebras inference engines
Scientific Accuracy: Nutrition data validated against USDA FoodData Central database
Voice-First Interaction: Natural voice input for hands-free recipe generation
Privacy-Conscious: HttpOnly cookies, token blacklisting, and configurable rate limits
Developer-Friendly: Modular domain-driven architecture with clean separation of concerns
Cloud-Native Ready: Docker containerization with Vercel serverless support