AI Engineer for SmartView Chatbot

We are developing SmartView, an AI/GIS platform for public-space and municipal asset observations. We already have a SaaS interface, a GIS specialist and existing SmartView data. For this first phase, we are not looking for a large autonomous decision-support system. We want a practical AI chatbot that can answer questions about the data already available in SmartView. The goal is simple: When a user asks a question about a location, detection, asset or dataset, the chatbot should retrieve the relevant available context, answer clearly, cite the source data and state what information is missing. This is not a generic customer-service chatbot, but it is also not a complex autonomous decision-making system. Scope of phase 1 You will build a working chatbot prototype that can: - connect to SmartView sample data through APIs, database queries or structured files; - retrieve context for a selected location, asset, detection or observation; - answer questions using only available data; - cite the source records or documents used in the answer; - distinguish between known facts, assumptions and missing information; - avoid unsupported conclusions; - return a simple structured response with answer, sources, confidence and missing information; - expose the chatbot through an API so it can later be integrated into our existing SaaS/GIS platform. Important boundaries The chatbot should not: - invent standards, measurements, causes or conclusions; - create automatic work orders; - make final operational decisions; - recommend measures unless the available source data clearly supports them; - interpret raw LiDAR, thermal or camera data directly unless structured outputs are already available; - replace human review; - overcomplicate the first version with unnecessary architecture, frameworks or multi-agent systems. How we want to work We are looking for someone who can move fast, think clearly and deliver working software without turning a focused first phase into a large consulting project. We are not looking for hour-chasers or people who try to create as many billable hours as possible. We value speed, ownership, clear communication and practical delivery. The first phase should be small, focused and testable. If the collaboration works well, we are looking for a long-term AI engineering partner who can continue improving the SmartView chatbot and later help expand it into more advanced features. We prefer someone who: - works quickly and independently; - gives realistic estimates; - avoids unnecessary complexity; - communicates clearly when something is blocked; - focuses on working output rather than long theoretical documents; - can explain technical choices in simple terms; - and wants to build a long-term relationship rather than maximise short-term hours. Preferred stack A practical stack may include: - Python; - FastAPI; - PostgreSQL/PostGIS; - pgvector or another vector search option; - OpenAI or Azure OpenAI; - simple RAG or lightweight agent workflow; - Docker; - structured JSON outputs. We are open to a practical technical proposal. We prefer a simple, reliable and maintainable first version over an over-engineered architecture. First deliverables The first phase should deliver: 1. A short architecture plan. 2. A working chatbot prototype using sample SmartView data. 3. Retrieval from structured data and documents. 4. Source references for answers. 5. Missing-information handling. 6. A simple structured JSON response format. 7. Basic test questions and expected answers. 8. Setup and handover documentation. Who we are looking for We are looking for a hands-on AI engineer who has experience with: - RAG or data-grounded chatbots; - Python and FastAPI; - database or API integrations; - structured outputs; - source references; - basic guardrails against unsupported answers; - clean documentation and handover. Experience with GIS, PostGIS, asset data, sensor data or municipal software is useful, but not required as the main skill. We already have GIS expertise available. Please apply with Please include: - one relevant chatbot, RAG or AI data assistant you personally built; - your exact role in that project; - the stack you used; - how you handled source references and missing information; - how you would approach this first SmartView chatbot prototype; - how you would keep the first phase simple and avoid unnecessary complexity; - your availability for a small paid test using sample data. Please do not send a generic AI-generated application. We are looking for someone practical who can build a working, evidence-based chatbot quickly and responsibly, without overcomplicating the first phase.