Sminex has developed an internal web application called Sminex Start to automate the onboarding process for new employees. The system replaces manual data entry from identity documents with an AI-driven approach, allowing for faster processing of HR documentation.
The platform utilizes a locally deployed vision-language model from the Qwen family to analyze and extract data from documents. To ensure accuracy, the company implemented a two-stage process: a dedicated OCR service extracts text from images, while the AI model organizes the information into structured fields.
The system includes an algorithmic module for address normalization that integrates with the FIAS database via a REST gateway. This ensures that addresses extracted from passports are validated and compatible with the 1C:ZUP corporate system, preventing errors in reporting to tax and social authorities.
Security is maintained by keeping all data processing within the company's internal perimeter. The AI model is hosted locally, meaning no documents are sent to public clouds. Once processed, data is transferred directly to corporate systems, and the application does not function as a long-term document archive.
The onboarding workflow is personalized based on the employee's role and legal entity. It includes digital signing of local regulations, photo capture for access control systems, and integration with corporate messengers to provide real-time support from HR staff during the process.
