AI Text Recognition Solution for European HR Start
One of the major challenges faced by SaaS companies is the ongoing need to enhance their features to maintain existing customers and attract new ones. For HURMA, the primary users of the product are recruiters. A common issue they encounter is the manual processing and transfer of data from numerous resumes in various formats. Additionally, there are other hurdles that complicate the recruiters' tasks: resumes can arrive in different formats such as PDF, DOC, or image files, and text formatting can vary significantly. For instance, names may appear in different orders, and users from multilingual regions may receive resumes in several languages. Given IT Svit's considerable experience in applying AI in business, our team leveraged this technology to create a solution. The goal was to develop a service that can recognize text from resumes and transfer the extracted data into HURMA's internal system. To realize this concept, our engineers evaluated multiple text recognition algorithms. While testing solutions like NER from NLTK, CoreNLP, and other off-the-shelf options, we found they were inadequate for identifying Slavic names. Ultimately, we chose a Google library solution combined with the Celery library to ensure stability during peak loads. As mentioned earlier, the varying formats, text patterns, and languages of the resumes required the model to be specifically trained for effective operation in such diverse conditions. This testing and refinement process lasted two months. Now, users can submit one or multiple resumes in formats such as PDF, TXT, DOC, and IMG, and receive standardized data in HURMA. Furthermore, when a candidate applies for a position on a job portal and includes a resume, the service automatically processes and inputs the data into the system. This solution enables HURMA users to save significant time and enhance their workflow.