This Python script reads USCIS AAO decision PDFs, extracts the main reasons for denial (for the 3 NIW prongs), and compares each case side-by-side with your own EB2-NIW petition strengths.

You can automatically compare your case against 7,800+ real AAO decisions!

• Scrape USCIS AAO PDF decision documents
• Use OpenAI GPT-4o model to summarize NIW prong failure reasons
• Compare each decision to your own petition
• Batch processing for thousands of PDFs
• Clean CSV file output for easy review

git clone https://github.com/JP2670/eb2NIW.git cd eb2NIW pip install -r requirements.txt

This script uses OpenAI GPT-4o model.
You need an OpenAI API key to use it.

• Signup for free at: https://platform.openai.com/signup
• After signup, go to API Keys page
• Click Create New Secret Key and copy it
• Paste your API key inside the Python script here:

openai.api_key = "your-api-key-here"

⚡ Important: You will need to add a small amount of balance ($5–$10) to your OpenAI account to run this full project.

Cost estimate:

• Around $30–$40 to process ~7,700 PDFs using GPT-4o-mini model (April 2025 prices).

Before running the script, you need to tell the program what your own EB2-NIW petition looks like.

You do this by editing the following part of process_eb2niw_prongs_compare.py:

In the beginning of the script, you will find this block:

my_case = { "Prong1": "Summarize why your National Importance is strong", "Prong2": "Summarize why you are Well Positioned", "Prong3": "Summarize why Labor Waiver is justified for you" }

You need to replace the example texts with short 1–2 sentence summaries based on your real EB2-NIW petition.

my_case = { "Prong1": "Strong: working on cutting-edge technology aligned with U.S. innovation goals.", "Prong2": "5+ years professional experience in major U.S. organizations, demonstrated leadership roles.", "Prong3": "Immediate contribution to national innovation efforts; delay would harm competitiveness." }

my_case = { "Prong1": "Strong: Research directly improves U.S. healthcare outcomes using AI for early disease detection.", "Prong2": "4+ years leading projects at a top U.S. university hospital and multiple published papers.", "Prong3": "Immediate healthcare application; delay would risk public health improvements." }

my_case = { "Prong1": "Strong: Directly advising U.S. state governments on economic policy initiatives.", "Prong2": "8 years of leadership roles in U.S. think tanks, direct policy impact proven.", "Prong3": "Delay would harm ongoing critical public sector projects; immediate national impact needed." }

• Be short (1–2 sentences max)
• Be direct: mention U.S. impact, leadership, urgency
• Match the style you would use in your NIW petition or recommendation letters
• No need to write full essays — this is just for automated comparison.

✅ This will make your script work correctly for your unique situation!

The script will generate a file called summary_prongs_comparison.csv with the following columns:

• The Master_file provided contains over 7,800 AAO decision PDF links (as of April 2025).
• Some non-EB2-NIW cases are mixed in (e.g., EB-1, EB-3) — filtering is a future update.
• The script automatically detects non-NIW cases and marks them.
• For 'mixed' verdict cases, you can manually check the PDF links to make your own detailed judgment.
• You can even use the PDF link with ChatGPT to create deeper comparisons with your petition if needed.

Pull requests, improvements, and feature ideas are welcome!

Open-source for personal and educational use.