The original title is: "Building MutemoOS: A jurisdiction-aware legal research system for Zimbabwean law firms"

Short summary

A developer built MutemoOS, an AI-assisted legal research system for a Zimbabwean law firm, addressing jurisdiction-specific challenges that generic RAG pipelines miss — including scattered case law, colloquial-to-statutory term mapping, and dual legal systems. The system uses a two-stage Claude synthesis pipeline (Haiku for grounding checks, Sonnet for constrained synthesis) with explicit source-sufficiency badges to prevent hallucinated citations. A two-layer fallback combining FTS and Zimbabwe-specific query expansion handles terminology gaps where embedding models fail on local legal concepts like lobola and small houses.

• •Built MutemoOS with FastAPI, ChromaDB, and Claude for Zimbabwe-specific legal research replacing a 118-matter notepad system
• •Two-stage synthesis pipeline (Haiku ground-check + Sonnet constrained synthesis) eliminates hallucinated citations with visible source-sufficiency badges
• •Two-layer fallback (FTS + Claude query expansion) solves embedding model failures on Zimbabwean colloquial-to-statutory term mappings

Generated with AI, which can make mistakes.