Expert AI Chatbot

Expert AI Chatbot — Python / React / WordPress PROJECT OVERVIEW We have built a custom AI chatbot deployed on WordPress client websites. The system connects to client data via a RAG (Retrieval-Augmented Generation) pipeline — indexing structured and unstructured content into a PostgreSQL + pgvector database — and serves responses through a React chat widget embedded as a custom PHP plugin. We are looking for an experienced fullstack developer to join on a long-term contract basis: maintaining the codebase, improving performance, extending features, and integrating new client deployments. --- TECH STACK Backend (Python API) - Python 3.11 — FastAPI + Uvicorn - LangChain — RAG orchestration - OpenAI SDK + Anthropic SDK (GPT / Claude) - PostgreSQL + pgvector - SQLAlchemy ORM Frontend (WordPress plugin) - React 18 + TypeScript - Vite bundler - Tailwind CSS - TanStack Query (React Query) - WordPress / PHP custom plugin --- WHAT YOU'LL DO - Maintain and evolve the RAG pipeline (chunking strategy, embedding updates, retrieval quality) - Backend API development and debugging (FastAPI endpoints, SSE streaming) - Improve and extend the React chat widget (UI flows, new features, mobile optimization) - Support new client deployments: data ingestion, WordPress plugin configuration - Monitor performance and cost (LLM token usage, latency, vector DB queries) PROFILE Required: - Strong Python with FastAPI, LangChain or similar RAG orchestration framework - React + TypeScript, able to build and maintain production-grade components - Experience with OpenAI or Anthropic APIs in production - PostgreSQL — pgvector knowledge is a strong plus Nice to have: - WordPress / PHP plugin development experience - Familiarity with embedding pipelines, chunking, and retrieval evaluation ENGAGEMENT - Long-term contract, remote, async-friendly - English or French communication - Part-time or full-time depending on workload HOW TO APPLY Please include in your proposal: 1. A brief description of a RAG or AI chatbot project you have shipped 2. Links to relevant GitHub repos or live demos (if public) 3. Your experience with pgvector or alternative vector stores 4. Your availability and preferred engagement format