AzureChef

Inspiration

I was inspired by the growing need for personalized cooking assistance that combines AI-powered recommendations with professional culinary knowledge. Many home cooks struggle with recipe discovery, meal planning, and understanding cooking techniques. I wanted to create an intelligent assistant that could not only find recipes but also provide historical context, nutritional insights, and professional cooking tips.

What it does

AzureChef is an AI-powered culinary assistant that:

Searches through an extensive recipe database using natural language queries
Provides detailed recipe instructions with ingredients and quantities
Offers AI-enhanced explanations with cooking tips and variations
Includes cultural and historical context about dishes
Supports voice interaction through text-to-speech functionality
Allows filtering by categories like appetizers, main dishes, desserts, etc.

How we built it

AzureChef combines multiple Azure AI services with Gemini AI:

Backend: Python with PyPDF2 for recipe extraction
Search: Azure AI Search with custom index schema
Storage: Azure Blob Storage for recipe documents
AI: Gemini for natural language processing and response generation
Voice: Azure Speech Service for text-to-speech conversion
Frontend: Gradio for the web interface with custom styling

Challenges we ran into

Azure OpenAI Limitations: As student accounts don't support Azure OpenAI, i creatively leveraged Gemini as a primary LLM, requiring extensive prompt engineering to achieve similar culinary analysis capabilities.
Recipe Parsing: Extracting structured data from PDF recipes with inconsistent formats required complex regex patterns and error handling.
Voice-Text Synchronization: Aligning Azure Speech Service's audio output with Gemini's dynamic responses demanded careful timing management.