Case Study
Kanun AI
Professional Legal Document Retrieval. A production-ready Retrieval-Augmented Generation (RAG) API for semantic search over legal documents. Fast vector search, explainable results, and built-in evaluation tools for Nepal's legal corpus.
RAG API FastAPI
Project Snapshot
Role: Full-stack developer — architecture, API development, data ingestion, and evaluation tooling.
Tech stack details are outlined below.
Overview
Built a production-ready legal document retrieval API that surfaces relevant precedents in seconds with transparent, explainable results. The platform supports legal researchers with reliable citations and audit-ready sourcing.
Technical highlight: Implemented Sentence Transformers for high-accuracy embeddings and PyMuPDF for efficient document parsing.
The Result
Created a production-ready API capable of explainable legal search across thousands of documents, cutting manual research time and improving retrieval confidence.
The Problem
Nepal's legal corpus is vast and difficult to search semantically using traditional keyword methods.
The Solution
Built a Retrieval-Augmented Generation pipeline using FastAPI and Qdrant for sub-second semantic search and ranked legal retrieval.
Tech Stack
Key Features
- Sub-second semantic search with vector similarity scoring.
- Explainable retrieval with citations and source excerpts.
- Evaluation tooling to monitor search quality over time.