OpenJustice
An open-source legal AI platform that makes legal reasoning transparent and usable for professionals.
Overview
OpenJustice is an open-source legal AI platform designed to make legal reasoning explainable, transparent, and accessible. It allows lawyers, researchers, and policymakers to explore case outcomes through structured dialog flows that simulate professional reasoning.
The platform combines modern language models with legal expertise to reason through claims such as constructive dismissal, contract disputes, and employment law while keeping every decision traceable.
OpenJustice is a no-code platform that mimicks how legal professionals reason through problems
Tech Stack
Core Platform
Language
TypeScript
Core development language across backend and frontend.
Backend Framework
NestJS
Modular server framework using Domain-Driven Design and microservice architecture.
AI & Reasoning Layer
AI Orchestration
Agentic RAG Pipeline
Retrieves legal materials, performs reasoning with citations and confidence scores.
LLM Integration
GPT, LLAMA APIs
Powers legal reasoning, parsing, and citation generation.
Dialog Flow
Start, Extractor, Switch, Reasoning, Outcome
Nodes represent workflow stages for structured reasoning.
Memory
Facts Dictionary
Acts as in-memory knowledge base for facts, citations, and conclusions.
Data & Storage
Vector Database
Qdrant (current), Pinecone (earlier)
Stores legal document embeddings and metadata for semantic retrieval.
ORM / Schema Management
Drizzle ORM
Schema definitions, migrations, and type-safe database access.
File / Blob Storage
Microsoft Azure
Stores user uploads and normalized document sets.
Reactive Programming
RxJS Observables
Handles streaming updates and real-time output (SSE).
Infrastructure & Deployment
Containerization
Docker, Docker Compose
Local development and deployment consistency.
Inter-service Communication
NestJS message patterns, HTTP
Connects reasoning and retrieval services.
Frontend & Security
Frontend Framework
React, Next.js
Frontend not confirmed, but consistent with TypeScript usage.
Security / Compliance
Role-based access, SOC2 Type 2
Protects sensitive legal data, ensures compliance.
My Contributions
Forward Deployed Engineer
June 2025 - Present- •In one month developed OpenJustice version 2 from scratch with my co-developer, creating a RAG agentic workflow with 100+ Canadian case law documents and logic that reflects a legal professional's
- •Generated responses that closely resemble a lawyer's by translating the reasoning process that lawyers use into programmable logic via 5 nodes: start node, fact node, reasoning node, switch node, and output node
- •Creating a 30 page comprehensive guidebook for venture capital and NSERC partners, as well as new OpenJustice employees, containing introductory (mission statement, values, problem scope) and system-level (architecture, RAG workflow, logic) information
- •Constantly improving OpenJustice through feedback from NSERC partners at top law firms and 30 law students, helping both groups navigate the platform and addressing the painpoints they encounter
Capstone Project
September 2024 - April 2025- •Collaborated with 3 other computer engineering students to resolve high priority GitHub issues, implementing TextGrad and feedback mechanisms
- •Read 10+ research papers on reinforcement learning with AI (RLAI) and completed the CourseEra course to understand industry best practices and promising research ideas
- •Improved response accuracy by 7% by developing an RLAI implementation script using open-source code and the proximal policy optimization algorithm
Explore more: Extracurriculars · Internships · Travels