Problem
A recruitment company specializing in IT hiring struggled with time-consuming CV processing, which largely relied on manually entering data into its internal recruitment system. CV documents varied widely in length (from one to a dozen-plus pages), formatting, and file type (DOCX, PDF).
As a result, manual data transcription was not only slow but also error-prone and limited how many candidates the team could handle.
An additional challenge was the need to translate CVs into different languages, anonymize personal data, and tailor the content to specific roles, which increased the workload even further. In some cases, preparing a single CV took up to 60 minutes.
The project goal was to significantly reduce processing time while improving consistency and data quality across the recruitment pipeline.
Wyzwanie
The biggest challenge was embedding the solution into an existing recruitment application built in legacy PHP.
Further challenges included the wide variety of CV formats, non-standard document layouts, and the need to maintain high data extraction quality.
Rozwiązanie
COGITA built an AI-based solution using modern multimodal models and LLMs capable of analyzing the content of PDF and DOCX documents.
The system automatically extracts all key candidate information from a CV and produces the output as a structured JSON file, enabling immediate integration with the client’s existing recruitment application.
In addition to data extraction, the solution included:
- CV translation
- a module for matching a candidate profile to a specific role
- a CV anonymization module (anonymizing personal data, company names, profile links, university names, etc.)
Importantly, the solution was deployed without requiring changes to other parts of the client’s infrastructure (the PHP application), so it delivered value from day one.
Wynik
The system reduced CV processing time by 90%, significantly accelerating candidate evaluation. In the first two months, it processed over 2,000 documents with 98% data extraction accuracy. At the same time, the consistency and quality of information in the recruitment pipeline improved, and the HR team could focus on analytical and relationship-driven work instead of administrative tasks.
