All projects
Overview
ALOud is a perfume and fragrance e-commerce platform built with ASP.NET Core Razor Pages and Entity Framework Core on SQL Server, with a Redis-backed shopping cart. I extended it with an AI layer so shoppers can describe what they want instead of browsing filters.
The assistant combines three techniques: semantic product search over a self-hosted embedding pipeline (Ollama and Qdrant), a ReAct agent that calls tools to manage the cart in real time, and an NRules engine that keeps recommendations deterministic and explainable. A separate Vue dashboard manages the back office.
What I built
- Architected a hybrid AI system combining RAG retrieval, a ReAct agent and a rule engine for commerce automation.
- Built a self-hosted embedding pipeline with Ollama and Qdrant for scalable semantic product search.
- Developed a multi-tool conversational agent enabling real-time cart management through Redis-backed state.
- Integrated a deterministic NRules engine for explainable climate and persona-based recommendations.
Design and architecture
- Retrieval: product descriptions are embedded locally with Ollama and stored in Qdrant, so search stays self-hosted and scales without a paid API.
- Agent: a ReAct loop selects tools such as search and cart actions, with cart state held in Redis between turns.
- Rules: climate and persona-based recommendations run through NRules, so each suggestion can be traced to a rule rather than a model guess.