Launching today

Doksima

Launching today

Document fraud review: demo & private beta waitlist

1 follower

Document fraud review: demo & private beta waitlist

1 follower

Explore how Doksima helps review suspicious French payslips and bank statements. Try a guided demo with fictional documents, explained comparisons and source PDFs—no account needed. The app is not open yet. Join the private beta waitlist.

Launch tags:Productivity•Fintech•Artificial Intelligence

Launch Team / Built With

OpenAI

GPT-6 Astra Challenge

How did Astra change the scope or ambition of what you built?

Maker

GPT-6 Astra in Codex helped us pursue Doksima as a coherent product, not just isolated document-checking experiments. The ambition is to help brokers investigate suspicious payslips and bank statements by connecting documents, tracing findings to their sources, and separating contradictions from signals that need review. Astra supported our development iteration. For this launch, Codex work included the bilingual product website, animated hero, routing fixes and deployment checks. We also use AI agents through OpenAI's API for development and R&D, checking their outputs against code and reproducible experiments. The boundary matters: an AI-generated explanation is not proof of fraud. Doksima is an early-stage prototype for French document workflows; experimental forgery-detection results are not universal accuracy claims, and the reviewer keeps the decision.

Report

2h ago

Maker

📌

Hi Product Hunt! I'm Billel, the founder of Doksima. A payslip can look convincing while its numbers contradict each other—or the bank statement submitted alongside it. I'm building Doksima to help brokers investigate those inconsistencies and follow the evidence back to the document. The product brings together: • Calculation and cross-document consistency checks. • Findings linked to the source information behind them. • Review notes, document versions and requests for replacement documents. • A distinction between a demonstrated contradiction, a suspicious signal and insufficient evidence. We used GPT-6 Astra in Codex to build and iterate on Doksima. For this launch, that work included our bilingual product website, animated hero and deployment checks. We also use AI agents through OpenAI's API to support development and R&D: writing and reviewing code, building tests and challenging detection hypotheses. An AI-generated explanation is not, by itself, proof of fraud. We're also researching detection of AI-regenerated documents and local edits. Those experimental results are not a promise that Doksima detects every forgery. Doksima is an early-stage product. The app and source documents are currently in French; the product website is available in English. The reviewer stays in control, and Doksima does not make lending decisions. I'd love feedback from brokers, document reviewers and fraud teams: what evidence would make a suspicious-document alert genuinely useful in your workflow?

Report

2d ago

OpenAI

GPT-6 Astra Challenge

How did Astra change the scope or ambition of what you built?

Maker

Report

2h ago

Maker

📌

Report

2d ago