~/portfolio/projects/immigration-document-intelligence-platform
← cd ../projects

#01 · 2024

Immigration Document Intelligence Platform

Platform build (pre-launch)
view code →
readme.md

Pre-launch AI platform · Python · FastAPI · RAG · LLMs

2

full system iterations

4-tier

role-based access control

47/47

tests passing

my_role: Founder & engineer - designed the legal taxonomy, built the RAG pipeline, FastAPI backend and multi-tenant platform end to end.

Immigration attorneys processing EB1 and EB2 visa applications handle hundreds of unstructured documents per case, each of which must be correctly classified against USCIS evidentiary criteria before it can be used in a petition. I built an end-to-end AI platform that automates this process using LLMs for document classification, a custom legal taxonomy aligned to USCIS criteria, and a RAG architecture for semantic document retrieval across case files. The system went through two full iterations, from a single-tenant prototype to a multi-tenant platform (pre-launch) with full data isolation per law firm, JWT authentication, four-tier role-based access control, and 47/47 passing tests. Self-reported; no usage or accuracy metrics yet.

## AI-powered document processing for EB1 & EB2 visa applications

A multi-tenant platform (pre-launch) that applies LLMs and RAG to automate document classification, extraction, and retrieval for U.S. immigration law firms - built in two iterations from prototype to enterprise architecture.

## What this project does

Immigration attorneys processing EB1 and EB2 visa applications deal with hundreds of unstructured documents per case - recommendation letters, published research, employment records, awards, and more. Each document must be correctly classified against USCIS evidentiary criteria before it can be used in a petition. This system automates that process using LLMs and Retrieval-Augmented Generation (RAG), turning a manual, error-prone workflow into an intelligent pipeline that:

  • ›Classifies uploaded documents against EB1/EB2 USCIS criteria automatically
  • ›Extracts structured data from unstructured PDFs and DOCX files
  • ›Retrieves the most relevant precedent documents for a given case using semantic search
  • ›Scales across multiple law firms with full data isolation per tenant
meta.json
year
2024
status
Platform build (pre-launch)
tags
LLM pipelinesLegal tech
gallery [1]
Immigration Document Intelligence Platform - figure 1