Hello HN,
I built Compare Contractor Quotes to solve a problem I faced during my own home renovation: trying to compare three different quotes that looked nothing alike. One was a detailed PDF, one was a Word doc, and one was essentially a photo of a napkin.
It was nearly impossible to see if they were quoting the same scope of work. I realized that the "cheapest" quote was actually just leaving out expensive items (like debris removal and permits) that would have hit me as change orders later.
What it does:
The tool takes unstructured contractor bids (PDFs, images, scans) and uses LLMs to parse them into a structured data format. It then:
Normalizes the data: Maps different terminologies to a standard scope of work.
Identifies Gaps: Highlights if one contractor included a line item that others missed.
Flags Risks: Checks for front-loaded payment schedules or vague language that puts you at risk.
How it works:
We use a combination of OCR and a fine-tuned LLM pipeline to extract line items, prices, and terms from the raw documents. The challenging part was getting the model to understand the context of construction lingo (e.g., differentiating between "rough-in" and "finish" plumbing) across different trades.
The Goal:
To give homeowners the leverage of a General Contractor. By standardizing the bids, you can see exactly where the price differences are coming from and negotiate from a position of data rather than guessing.
It’s live now at
https://comparecontractorquotes.com.
There is a free preview mode so you can test it out without paying. I’d love to hear your feedback on the extraction accuracy and if there are specific "red flags" you think I should add to the detection logic!