RestAI — AI-Powered Restaurant Management SaaS
github.com/ashour-git/Res... | Python, FastAPI, Next.js, LightGBM, Groq Llama 3.3 70B, RAG, pgvector, Docker,
PostgreSQL
• Architected a full-stack SaaS platform (FastAPI backend, Next.js frontend, separate ML service) with JWT
authentication and role-based access control, backed by a pytest test suite covering API, forecasting, and
recommender logic.
• Implemented LightGBM demand forecasting with Optuna hyperparameter optimization and lag/rolling window
features for time-series prediction.
• Built a RAG-based AI assistant using Groq Llama 3.3 70B and pgvector for semantic retrieval over menu and
operational data, with a hybrid co-occurrence and content-based recommender system