Agentic AI Engineer — Text-to-SQL Analytics Chatbot (LangChain/LangGraph + OpenAI)

We're building an **agentic AI, text-to-SQL analytics chatbot**. It lets users ask questions about their business data in plain English — things like _"Show me the revenue trend,"_ _"Which source brought the most leads last month?,"_ or _"Where are we losing people in the funnel?"_ — and get clear, accurate, data-backed answers. ​ Behind the chatbot is a **large relational database (30+ tables)** covering areas like leads, appointments, revenue, and marketing performance. The AI agent understands each question, figures out where the answer lives across those tables, and responds with the numbers, trends, comparisons, and breakdowns the user asked for — always grounded in real data, never made up. ​ We're looking for an **AI engineer** who is strong on LLM / agent orchestration and text-to-SQL to help us build out and improve this agent. ​ ### What you'll be working on ​ - Build and improve an agentic AI that turns natural-language questions into accurate answers over a 30+ table database. - Support analytics use cases end to end: single-period metrics, time-series trends, period-over-period comparisons, grouped breakdowns, and funnel / diagnostic questions. - Keep answers accurate and trustworthy — the agent must return real numbers from the data and never fabricate them. - Improve prompts, accuracy, and the range of questions the agent can confidently handle. ​ ### Tech stack (you should be comfortable with most of this) ​ `Python` · `FastAPI` · `LangChain` & `LangGraph` · `OpenAI API` · `PostgreSQL` · `Pydantic` · (`LangSmith`, `Docker`, `DigitalOcean` a plus) ​ ### Required skills ​ - **Agentic AI / LLM orchestration** (LangChain / LangGraph or similar) and tool / function calling. - **Text-to-SQL** / natural-language-to-SQL over relational databases. - Strong **Python** and backend API development (FastAPI). - Solid **SQL** and **PostgreSQL**. - A focus on **accuracy** — making sure the AI returns correct, non-hallucinated answers. ​ ### Nice to have ​ - Prompt engineering and LLM evaluation / observability experience. - Marketing / sales analytics domain knowledge (leads, appointments, revenue, paid ads). - Docker / cloud deployment experience. ​ ### Deliverables / definition of done ​ - Working, tested improvements to the agent. - Accurate answers verified against real data across the supported question types. - Updated documentation where relevant. ​