pls feel free to report any issue

*Building an AI-Powered Document Understanding Tool – Feedback Welcome*

Hi HN!

I'm working on a tool called *ClearDoc*, which uses AI to extract structured data from unstructured documents like invoices, bills of lading, and certificates. The biggest challenge we've faced so far is accurately extracting data from complex documents, especially those with tables and nested fields.

### What I’m Looking to Discuss: - How do you approach extracting data from complex documents like invoices or contracts? - If you’ve worked with OCR or document processing tools, what have been your biggest challenges?

We’ve built a demo that uses PaddleOCR and LLMs to extract and align data. I’d love to get your thoughts on how we could improve the accuracy of data extraction, or whether you think a no-template approach is valuable.

If you’re interested, feel free to try out the demo (no sign-up required) and let me know your thoughts!

[ClearDoc Demo](https://cleardoc.v5ent.com/)

Looking forward to your feedback!

#AI #OCR #MachineLearning #DocumentProcessing

hi HN! — just pushed a new update to *ClearDoc*, my AI tool to extract *structured data from unstructured documents* (like invoices, logistics forms, certificates, etc.)

---

### What’s New:

*HTTPS Enabled:* The live demo is now secure at [https://cleardoc.v5ent.com](https://cleardoc.v5ent.com), so no more browser warnings.

*Improved Homepage Messaging:* Based on user feedback, the homepage now has a much clearer value proposition and simplified CTA. For example, “Reasoning Output” is now simply “View Extracted Data.”

*Performance Tweaks:* Faster processing, better alignment, and cleaner output.

*Coming soon: Confidence Scores + Feedback Loop* So users will see which extracted fields the AI is “most sure” about — and be able to correct any errors to improve future results.

---

### What is ClearDoc?

ClearDoc helps you *turn messy PDFs/images into clean JSON* — without templates, without fine-tuning, and fully self-hostable.

It combines: - OCR (PaddleOCR) - LLM (OpenAI-compatible) - Field alignment + visual overlays - JSON Schema output (customizable)

Demo: https://cleardoc.v5ent.com Video: https://www.youtube.com/watch?v=u83T6iewfNs

---

### Who is this for?

- Developers building document-based tools - Finance / accounting teams who copy-paste data - Logistics / trade teams processing paperwork - Anyone who hates manually parsing PDFs

---

### I'm looking for:

1. Early users with real docs they want to process 2. Edge cases you'd like to see it handle 3. Feedback on the extraction quality / experience

I’d love to hear what you think — or help if you're facing similar problems.

Thanks — Charles