Local AI Engineering with Ollama #17: Universal Intelligence Engine (Three-Pass AI System) Part 1

🚀 Local AI Engineering with Ollama #17: Universal Intelligence Engine Welcome to the most advanced project in the Local AI Engineering with Ollama series so far. In this lesson we build a Universal Intelligence Engine capable of analysing any website and automatically generating a domain-specific intelligence report. Unlike traditional AI summarisers that use one prompt for every website, this system first determines what type of organisation it is analysing before selecting the most appropriate intelligence framework. The result is dramatically better reports with lower token consumption and higher relevance. Everything runs locally using Ollama. No cloud APIs. No external inference services. No data leaves your machine. What You'll Learn ✅ Three-Pass AI Architectures ✅ Domain Classification ✅ Dynamic Prompt Selection ✅ AI-Powered Link Discovery ✅ JSON-Constrained Outputs ✅ Domain-Specific Intelligence Reports ✅ Multi-Page Content Aggregation ✅ Prompt Registry Design ✅ Temperature Strategy per Pass ✅ Enterprise AI Patterns ✅ Local AI Governance ✅ Regulated Industry Deployments The Three-Pass Architecture Pass 1 — Domain Classification The model determines the organisation type: Healthcare Financial Services University Government Technology NGO Other This classification determines everything that follows. Pass 2 — Smart Link Discovery The model analyses website navigation and selects the most relevant pages for the detected domain. Examples: Healthcare → Services, Quality, Research University → Research, Rankings, Partnerships Financial → Investors, Products, Strategy Technology → Products, Platform, Market Output is returned as structured JSON. Pass 3 — Domain-Tailored Intelligence Report The system automatically selects the correct intelligence framework and generates a detailed report. Examples: 🏥 Healthcare → Clinical Services, CQC, Governance 🏦 Financial → Investment Considerations 🎓 University → Research Excellence & Industry Partnerships 💻 Technology → Product Strategy & Market Position 🏛 Government → Policies & Accountability 🌍 NGO → Mission, Impact & Funding Why This Matters Most AI applications use a single prompt. Production AI systems use orchestration. This project demonstrates: AI-driven classification AI-driven navigation Dynamic prompt routing Domain-aware reporting Multi-step reasoning pipelines These are the foundations of: ✅ AI Agents ✅ Research Agents ✅ Competitive Intelligence Systems ✅ Enterprise AI Platforms ✅ Multi-Agent Architectures Domains Supported Healthcare NHS Trusts Hospitals Healthcare Providers Financial Services Banks Insurance Investment Firms Universities Research Institutions Higher Education Government Public Sector Organisations Technology Software Companies SaaS Providers NGO / Charity Non-Profit Organisations Other Generic Intelligence Reports Technologies Covered Ollama Python OpenAI SDK BeautifulSoup JSON Outputs Dynamic Prompt Routing Local LLMs Llama 3.2 Mistral Multi-Pass AI Systems Coming Next ▶️ Universal Intelligence Engine Demo ▶️ Conversational AI with Memory ▶️ Sliding Window Memory ▶️ Context Management ▶️ AI Agents ▶️ Tool Calling ▶️ Multi-Agent Systems ▶️ Production AI Architecture 👨‍💻 Ahmed Mahmoud AI Solution Architect | Snowflake Solution Architect | Principal Data Engineer Helping organisations move from AI demos to production-ready AI systems.