Litigma is a GDPR-aware legal document workflow platform built to ingest sensitive documents, extract and process their contents, and support workflow-driven AI assistance across a full-stack application.
This repository is a public-safe case study describing the system, the architecture, and the engineering work I did on it. It is not the full private source repository.
The platform combines:
- a React/Vite frontend for document and workflow UI
- Rust/Axum backend services for APIs and workflow orchestration
- Postgres-backed workflow state and async job tracking
- OCR and extraction workflows for document processing
- async LLM-driven tasks for structured workflow outputs
- health checks, metrics, and Docker-based local development
Document-heavy legal workflows often involve:
- large numbers of uploaded files
- OCR and extraction from inconsistent document formats
- structured workflow steps that depend on document content
- strict handling of sensitive personal data
- the need for auditable outputs rather than free-form AI responses
Litigma was built around those constraints.
My work focused on engineering across backend and workflow-heavy parts of the stack.
Areas I worked in:
- Rust/Axum backend services
- async worker and workflow behavior
- Postgres-backed state and queue-driven processing
- API contracts and workflow execution logic
- health checks, metrics, and local Docker development flows
- integration points between backend systems and user-facing workflow features
- Rust
- Axum
- React
- TypeScript
- Postgres
- Docker
- OCR and extraction pipelines
- async LLM-assisted workflows
React/Vite application for case, document, and workflow interaction.
Rust/Axum service handling authenticated routes, workflow APIs, and system orchestration.
Async workers for OCR, extraction, and LLM-assisted workflow tasks.
Postgres-backed storage for workflow state, job tracking, and supporting application data.
Health endpoints, metrics exposure, and Docker-based local development for repeatable setup and debugging.
- keeping workflow state consistent across async processing steps
- turning extracted document data into structured downstream outputs
- making complex system behavior observable and debuggable
- exposing workflows clearly in the product without losing auditability
- working across API, processing, and UI boundaries in one system
This project is the clearest example of the kind of software work I want to keep doing:
- backend-heavy engineering
- workflow systems
- practical AI product integration
- software where correctness and maintainability matter
If I were continuing this public case-study repo, I would add:
- an architecture diagram
- screenshots of the workflow UI
- a sample request/response section for key APIs
- a walkthrough of one workflow end to end
- a short section on testing strategy and reliability concerns
This repo is intentionally public-safe. It summarizes the architecture and engineering work without exposing sensitive implementation details, internal configuration, or private data.