mise.ai

An AI-powered restaurant analytics platform that helps reduce food waste, optimize ordering, and make data-driven decisions through ML predictions, LLM-powered insights, and interactive dashboards.

mise.ai is a restaurant analytics platform developed as a team project for CS4505 Software Architecture during my M.Sc. in Computer Science at TU Delft. The system helps restaurants reduce food waste, optimize ordering, and make data-driven decisions.

1. Product & architecture

   • React frontend communicates with a FastAPI core server over REST

   • Core server persists data in PostgreSQL and delegates ML/LLM work to a dedicated ML server

   • Async tasks (predictions, insight generation) are enqueued via RabbitMQ

   • Workers process jobs and write results back through Redis and result queues

   • Kubernetes deployment with KEDA auto-scales ML and LLM workers from 0–10 replicas based on queue depth

2. Key features

   • Interactive dashboards with sales metrics, wastage analysis, and weather impact charts

   • Drag-and-drop CSV/Excel/image (OCR) data ingestion with duplicate detection

   • Random Forest models for food wastage and per-item order quantity predictions

   • LLM-powered natural-language Q&A and auto-generated insight reports

   • Asynchronous PDF report generation and email delivery

3. Team contribution

   • Full-stack development across the React/TypeScript frontend and FastAPI backend services

   • Service integration for async prediction and LLM pipelines

   • Kubernetes deployment configuration and microservices orchestration