Legibus — Legal IA (LLM + RAG)

AI-powered legal solution combining LLM and RAG, designed to provide contextualized and verifiable answers.

legal ai langchain django rag

Legibus

Legibus is an AI-powered legal assistant project.
It was designed around a strong constraint:

Not to replace humans. To augment them.

Vision

In the legal field, errors are not acceptable.
Legibus was therefore designed with a radically different approach from mainstream LLM tools:

  • No black-box logic
  • No opaque answers
  • No unrealistic promises

Architecture

The system is built around:

  • LLMs orchestrated through controlled pipelines
  • A RAG layer for document grounding
  • A dedicated Django backend
  • Control APIs
  • A strict separation of concerns: ingestion, indexing, generation

My role

On Legibus, I worked as:

  • Technical architect
  • AI integrator
  • Infrastructure lead
  • Backend orchestrator

I collaborated with a PhD in AI and an R&D team in a co-construction mindset.

Technical principles

  • Explainable logging
  • Data versioning
  • Controlled embeddings
  • Environment isolation
  • Response monitoring

Legibus is not a magic product.
It is a serious tool for a serious domain.

AI should not replace.
It should illuminate.