Artificialy supported DeepCloud in the development of DeepLegal, a platform based on an Agentic RAG (Retrieval-Augmented Generation) architecture that transforms the way lawyers conduct legal research.

Key components of the solution:

• Knowledge base curated from 800,000 documents. A massive database querying all official Swiss laws and rulings from 2000 onwards, with over 10 million structured metadata entries to ensure precise and contextualised searches;

• LLM optimised for the legal context. A language model fine-tuned specifically for legal use cases, trained to understand legal terminology, argumentative structures, and regulatory logic;

• Parallel multi-agent pipeline. Rather than a single processing flow, DeepLegal employs a complex architecture in which several specialised AI agents work in parallel: a query expansion agent that analyses the user's question and breaks it down into sub-searches; parallel retrieval agents, one focused on legislation and the other on case law; a citation validation agent, which filters out irrelevant references and verifies source accuracy; and finally, a synthesis agent, which aggregates results from the various agents and generates a structured response.

• Zero data movement & on-premises deployment. The entire system runs locally on infrastructure hosted in Switzerland, ensuring that sensitive data is never moved or exposed to external jurisdictions.