Project Overview

Nexa AI: Transforming Legal Rental Assistance with Intelligent RAG Technology

Beyond Himalaya Tech successfully developed and deployed Nexa AI, an intelligent legal assistant that revolutionizes how tenants, landlords, and property managers access information about residential tenancy law in the Australian Capital Territory (ACT). Built on cutting-edge Retrieval-Augmented Generation (RAG) technology, Nexa AI provides instant, accurate, and cited responses to complex legal questions based on the official ACT Residential Tenancies Act 1997.

Industry: Nexa Ai Transforming Legal Rental Assistance With Intelligent Rag TechnologyDuration: 12-16 weeks
View Project↗
8
core tools used
4
delivery pillars
15
key improvements mapped
Live
launch status
The Challenge

Real Estate Customer Enquiry - Problem Statement.

Tenants, landlords, and property managers in the Australia face significant challenges. Some of the challenges are listed below:

Time-Consuming Research

Finding specific information in the 200+ page Residential Tenancies Act required extensive manual searching.

Inconsistent Information

Different sources provided conflicting interpretations of legal requirements.

High Consultation Costs

Legal advice for routine tenancy questions was expensive and often unnecessary.

Accessibility Barriers

Complex legal language made the Act difficult for non-lawyers to understand.

Outdated Information

Static resources couldn’t keep pace with legislative updates.

The Solution: Nexa AI Platform

One Smart Platform for Every Legal Stakeholder - Now for Real estate.

Nexa AI serves as a comprehensive legal assistant platform designed specifically for the Australian Capital Territory residential tenancy ecosystem, providing tailored experiences for each user type while maintaining a unified knowledge base grounded in ACT legislation.

Technical Architecture

Beyond Himalaya Tech designed and implemented a sophisticated AI-powered legal assistant using state-of-the-art technologies.

Backend Infrastructure

The backend was built with FastAPI in Python, ensuring high-performance API development and seamless scalability. The AI engine leveraged LlamaIndex integrated with OpenAI’s GPT-4, enabling intelligent document understanding and contextual reasoning. A PostgreSQL database enhanced with the pgvector extension powered semantic search, using OpenAI’s text-embedding-3-large model for precise data representation. To further improve accuracy, Cohere’s reranking API refined search relevance. Secure user authentication and management were handled through Supabase, maintaining both reliability and compliance.

Frontend Infrastructure

The frontend was developed using React 18 with TypeScript, providing a robust and type-safe environment for building scalable interfaces. The design combined Radix UI components with Tailwind CSS, resulting in a clean, accessible, and modern user experience. React Query managed data fetching and synchronization efficiently, while Supabase Auth supported seamless authentication, including social logins. The application was deployed on AWS App Runner, ensuring reliable performance and effortless scalability in production.

Core Features and System Capabilities

Key Features That Power Nexa AI

Nexa AI combines intelligent document processing, advanced RAG-powered retrieval, and a seamless multi-user experience, all built on enterprise-grade security. It delivers accurate legal guidance, role-specific insights, and real-time responses, ensuring scalable, secure, and user-friendly access for tenants, landlords, and property managers.

Intelligent Document Processing

Nexa AI automates the parsing of the ACT Residential Tenancies Act 1997, transforming dense legal text into searchable, structured data. Through semantic chunking of 900-character segments and hierarchical section mapping, the system ensures precise legal citations. Enriched metadata improves discoverability, while ACT-specific terminology and context understanding enable highly accurate, localized legal insights.

Advanced RAG Implementation

The platform employs a robust Retrieval-Augmented Generation (RAG) pipeline that combines vector similarity search with intelligent reranking. Using pgvector, Nexa AI retrieves the top 12 relevant sections before Cohere’s reranking identifies the best five matches. A custom citation synthesizer ensures every answer includes accurate legal references, while context-aware generation keeps responses relevant and compliant.

Multi-User Experience Excellence

Built to serve tenants, landlords, and property managers, Nexa AI offers a unified platform with role-based interfaces for personalized experiences. Users benefit from real-time chat, persistent history, and smart suggestions tailored to ACT law. The mobile-first design ensures accessibility from any device, while dark and light themes enhance usability and comfort.

Enterprise-Grade Security & Compliance

Security and compliance are central to Nexa AI’s architecture. The platform features authenticated user sessions, secure APIs, and adherence to GDPR and the Australian Privacy Act. With audit logging, role-based access controls, and strict privacy measures, every interaction remains safe, transparent, and accountable.

Tech Stack
Tools We Used

Frontend

React 18, Tailwind, Supabase Auth

Backend

FastAPI, Python, PostgreSQL with pgvector, Supabase

AI Engine

LlamaIndex + OpenAI GPT-4

Deployment

AWS App Runner for seamless cloud scalability