RATIN

What the company appears to be: RATIN is a self-reported proof-of-concept tool that converts mixed-format procurement documents (PDFs, HWP, XLSX) into an evidence-linked decision brief with exact navigation to source content. It was built as part of an OpenAI hackathon and is described as a dual-pane interface with PDF evidence stream and structured AI interpretation.

What changed: The project description states that RATIN was extended during a Build Week hackathon, adding features like exact page/text/spreadsheet-cell navigation, multi-target evidence handling, and GPT-5.6 integration with strict output contracts. It also mentions prior proprietary work on document normalization and provenance.

Single most important open question: Is there any evidence of real-world adoption or traction beyond the hackathon demo? The description contains no information about revenue, customers, usage, or commercial deployment.

The description states that RATIN:

• Converts mixed-format procurement documents (PDFs, HWP, XLSX) into an evidence-linked decision brief
• Uses a dual A4 layout interface with bid metadata on the left and complete Evidence PDF stream on the right
• Enables exact navigation to source content: PDF page, text bounding box, or spreadsheet cell location
• Connects logical findings back to physical source locations through "Evidence links"
• Implements deterministic document processing (normalization, identity, provenance) separate from model-based interpretation
• Uses GPT-5.6 through a strict structured-output contract for interpretation, not source generation

The product is described as a proof-of-concept built during a hackathon with no revenue or customer data.

The description states that RATIN was built to "close the gap" between AI-generated summaries and verifiable source evidence in procurement decisions. It positions itself as:

• Not just summarizing documents, but connecting every decision-relevant finding back to exact source evidence
• A tool for procurement decision-making that addresses information scattered across multiple document formats
• An approach that separates "evidence" from "interpretation" to maintain trustworthiness

The claim evolution shows a progression from a basic idea ("every AI-generated finding should be directly verifiable") to a specific technical implementation (dual-pane interface, exact navigation) and finally to a principle ("reliable document intelligence requires strict boundary between evidence and interpretation").

The description states that RATIN is designed for procurement decision-making contexts where:

• Information is scattered across PDFs, HWP documents, spreadsheets, public notices, and external reference data
• Users need to verify where every important statement came from
• Procurement decisions depend on information from mixed-format documents
• The tool supports bid metadata, participation requirements, cost references, conflicts, risks, questions for the issuer, and required actions

The target customer is not explicitly named but appears to be procurement professionals or decision-makers working with mixed-format procurement packages.

Not evidenced. The description contains no information about pricing, revenue streams, or business model.

The description states that RATIN:

• Separates deterministic document processing from model-based interpretation
• Owns deterministic layer responsibilities including: mixed-format document normalization, source identity and provenance, PDF page and bounding-box binding, XLSX cell binding, quantities, prices, and arithmetic, exact Evidence navigation
• Uses GPT-5.6 through a strict structured-output contract that does not invent source coordinates, prices, quantities, or document facts
• Implements exact PDF page and bounding-box navigation
• Implements exact XLSX cell navigation
• Supports multi-target evidence handling
• Has fail-closed live, cached, and fixture modes
• Includes automated tests (18/18 passed)
• Works without an API key in a validated frozen cache mode
• Has browser-level Evidence-jump validation

Not evidenced. The description contains no information about revenue, customers, usage metrics, or commercial deployment beyond the hackathon demo.

Not evidenced. The description contains no information about competitors, market positioning, or competitive landscape.

The description states:

• RATIN is a proof-of-concept built during a hackathon with no revenue or customer data
• The proprietary canonical pipeline is not included in the public repository
• The project has only one team member (SD Chae)
• Prior controlled validation results were from GPT-5.4, not the current GPT-5.6 implementation
• The final validated public build contains 21/21 logical links resolving to 26 exact physical targets with 0 fallback jumps, but this was a limited benchmark

Red flags include lack of commercial traction, single-person team, and no evidence of real-world adoption or revenue.

1. What is the actual market need for this specific type of procurement document navigation?
2. How does RATIN compare to existing procurement decision support tools in the market?
3. What are the actual costs of implementing and maintaining the deterministic document processing layer?
4. Has there been any real-world testing beyond the hackathon demo?
5. What is the path to commercialization or product-market fit?
6. How does the team plan to scale beyond a single developer?
7. What are the technical limitations of the current approach that would prevent broader adoption?