Yes. For Week 1, your goal is not “learn document AI deeply.” Your goal is:

Build a small but real intake system where one document becomes validated product data.

Do PDF first. Add email text as a small fallback. Leave image upload as stretch.

1. What should you work on: PDF, email, or image?

Main input: PDF claim/warranty document

Pick one document type only:

Best choice for you: Warranty / Insurance Claim Intake PDF

Why PDF first:

• It matches real document-AI companies better than plain email.
• Gemini supports native PDF document understanding and can extract into structured outputs. Google’s docs say Gemini can process PDFs with native vision, understand text/images/tables, and extract into structured output formats. (Google AI for Developers)
• It gives you layout, messy fields, dates, amounts, and missing-data problems.
• It prepares you for Week 3 RAG over policy documents.

Secondary input: paste email text

Add a textarea called “Paste email text instead” only if PDF upload works.

Do not do image upload this week unless Day 6 is already comfortable. Image support will drag you into OCR concerns too early.

Your Week 1 supported inputs should be:

MVP:
1. Upload PDF
2. OR paste email text

Stretch:
3. Upload image

2. Which AI model/API should you use?