Fanktank.ch Website
Challenge
Create a portfolio site that doesn't just describe AI capabilities, but demonstrates cutting-edge AI engineering through a production-grade conversational system.
Solution
Built a sophisticated multi-agent RAG system featuring:
- 6-agent orchestration pipeline that dynamically routes, researches, and synthesizes responses
- Hybrid search architecture combining semantic vectors, BM25 keyword matching, and Cohere reranking
- Real-time process visualization showing users exactly how the AI thinks through their questions
- Context-aware retrieval with purpose-tagged content and section-based filtering
- Parallel processing with dynamic query generation (1-5 queries based on complexity)
- Multiple LLM strategies using GPT-5's reasoning_effort parameter for optimal performance
Result
A production-grade demonstration of advanced AI engineering. The system intelligently processes ~750-word content chunks across 100+ documents, provides transparent insight into AI decision-making through live process visualization, and showcases how multiple specialized agents can work together to understand and answer complex questions. This isn't just a chatbot—it's a masterclass in modern RAG architecture.
Key Highlights
Multi-Agent Orchestration
6 specialized agents working in concert: Router, Search Specialist, Fact Extractor, Briefing Officer, and more.
Hybrid Search + Reranking
Semantic vectors + BM25 keywords + Cohere reranking for state-of-the-art retrieval accuracy.
Real-Time Process Visualization
Users see live updates as the AI routes, searches, distills, and synthesizes—full transparency into AI thinking.
