In a digital-era recruiting environment, the regular Applicant Tracking Systems (ATS) are not very useful in filtering through the large number of resumes received since they highly depend on the logic of key-words. This usually results in good applicants being lost in the cracks when they fail to execute the right phrases the system is searching and the system also attracts irrelevant applicants using the hopping feature by key words. Therefore, I have been considering JOBSY, which is an AI-based recruiting solution that attempts to resolve the problem by combining lexical and semantic matching to improve the efficiency and equity of hiring. JOBSY is implemented on top of NLP tricks it uses TF-IDF to perform the basic filtering of the key words and BERT to get a clearer understanding of the general meaning. It even includes a custom NER based parser that extracts structured information out of resumes, which can be used to better match the profiles of the candidate with job ads. All this has been developed using React.js and Flask which implies that the platform can manage real time and not hit its limit. Early tests indicate that JOBSY extracts an unbelievable amount of relevant candidates and the accuracy and quality of overall matches are better to resume screens compared to what the old ATS kids are accustomed to. Abstract: Resume Parsing is a technique that transforms resumes into a format compatible with AI or other machine processing tools to develop job vacancy matching, controlled by machine-executed instructions instead of human ones. TF-IDF or BERT along with Named Entity Recognition: This has been incorporated into the list to facilitate job matching by relying on instructions to execute tasks automatically, leading to an Applicant Tracking System rather than relying on human-executed instructions. < Semantic Embeddings Resume Parsing: This is included on the list to help facilitate job matching by using instructions to execute tasks

The JOBSY system addresses the weaknesses of the normal Applicant Tracking Systems through the application of the NLP methods and a hybrid matching engine that combines both the lexical and semantic analysis. JOBSY improves the performance of a resume matching a job in that it uses TF-IDF to ensure exact matching of keywords and BERT to gain more information about the job. It fully relies on custom NER parser, real-time responses, and its modular structure combined will ensure the accuracy of the candidate profiles along with fair filtering when dozens of applicants invade. The experiments reveal that the hybrid model implemented by it reduces false positives and negatives compared to ancient systems that were based on key words only in JOBSY. Its design developed using React.js in the front and Flask in the back allows the resume to be read and integrated asynchronously, allowing the recruiters to have the ranked candidates in sight rather fast. The fact that structured data and contextual embeddings were added, as well as an interface that is easy to use, makes the entire process of recruiting significantly more efficient and transparent. To make the decisions of the recruiters more trusted and transparent in the future, Explainable AI tools such as SHAP or LIME can be added to the work. The addition of domain relevant knowledge graphs may restrict semantic matching by establishing hierarchy among similar skills. And above all, the bias-auditing module should be considered in order to make the AI moral, addressing the gender, demographic, or age bias in the recruitment. All these upgrades will contribute to a better, more comprehensive, and smart recruitment system.

[1]C. Chandak, A. Kaushik, and P. K. Mishra, A Hybrid Model of Resume-Job Description Matching RF-IDF and BERT, International Journal of Research in Applied Science and Engineering Technology (IJRASET), vol. 12, no. 3. pp. 302- 308, 2024. I recently came across the article of Chandak, Kaushik, and Mishra, A Hybrid Model of Resume-Job Description Matching by TF-IDF and Bert (IJRASET, vol. 12, no. 3, 2024, p. 302-308). [2]S. Sougandh, R. Ranjitha and S. Sangeetha, A Review Paper on Resume Parser using AI, International Journal on Innovative Research in Technology (IJIRT) vol. 10 no. 7, pp. 112-115, 2023. A review by Sougandh, Ranjitha and Sangeetha of resume parsers using AI is published in IJIRT (vol 10, no 7, 2023, pp 112115). [3]R. Thangaramya, P. Vijayakumar and S. K. Lakshmanan, Resume2Vec: Intelligent Resume Embeddings to transform Applicant Tracking Systems, Electronics, vol. 14 no. 4, 2025. In Electronics, vol., issue 14, no. 4 (2025) the paper Resume 2 Vec was published by Thangaramya, Vijayakumar and Lakshmanan. [4]Essentially, it involves uses of the BERT model to calculate more dependable results in resume analysis and job assignment. <|human|>Essentially, it comprises applications of the BERT model in executing more reliable outcomes of analysis of resumes and allocation of employment. In one of the papers in the 2025 issue of IEEE Xplore (pp portions), Swapna, Lavanya, and Rani elaborate on the BERT-based resume analysis (p.16). [5]T. Sanyal and R. Roy, NER -based Resume Information Extraction to Get Better filtering of candidates, International Journal of Computer Applications, vol. 181, no. 12, pp. 33-39, 2023. At Int. Journal of Computer Applications (vol.181, no.12, 2023, p. 33-39) it is a resume extraction study with NER authored by Sanyal and Roy. [6]Prompt of engagement Research showing how vision-enhanced NER models can be used to layout-awarely process resumes is presented in [6]. The vision enhanced resume parser developed by Verma and Kumar was introduced at the IJCAI 2024 (pp 761768). [7]R. Deshmukh and A. Raut, TF-IDF Based Resume-Job matching: A Comparative Study, International journal of Advanced research in Engineering science and Management, vol. 6, no. 2, 2025. In the article by Deshmukh and Raut, comparing TF -IDF techniques are described (2025, IJARSEM, vol. 6, no. 2). [8]V. Singh, R. Sharma and M. Joshi, “Scalable Web Architecture Design of Resume Processing using Flask and React,” IJERT, vol. 11, no. 1, pp. 101-106, 2024. A system described by Singh, Sharma, and Joshi in the article in IJERT (vol 11, no 1, 2024, p.101-106) is a Flask/React resume system. [9]A. Kumar and V. Tripathi, A Survey of Techniques, Challenges and Future Directions of AI in Recruitment, International Journal of Artificial Intelligence Research, vol. 9, no. 3, pp. 123-132, 2025. The AIin-recruitment survey projected by Kumar and Tripathi is presented in IJAI Research (vol 9, no 3, pp123132). [10]S. Jadhav, M. Kale, and N. Khare, Contextual Resume Ranking with Feedback-driven Gossip Resume Embedding Adjustment, International Journal of Computer Science Trends and Technology (IJCST), vol. 11, no. 2, pp. 75-80, 2024. The context understands the ranking paper by Jadhav, Kale and Khare in IJCST ( vol 11 no 2, 2024 pp 75-80). [11]D. Devlin, M. Chang, K. Lee and K. Toutanova, BERT: Pre-training of Deep Bidirectional Transformer, Language Understanding, arXiv preprint arXiv:1810.04805 2019. It is the renowned BERT paper by Devlin et al. that is on arXiv (arXiv:1810.04805, 2019). [12]N. Reimers and I. Gurevych, SentenceBERT: Sentence Embeddings with Siamese Siamese BERT |networks, EMNLP, pp. 3982-3992, 2019. Sentence-bert was introduced at EMNLP 2019 by Reimers and Gurevych (pp39823992). [13]Unlike, of course, validity, reliability, and generalizability checks in experiments, the internal consistency within a study reliably represents descriptive psychometric integrity but aids the analysis of items that are susceptible to error (APA, 2019).<|human|>Still, in comparison to, naturally, validity, reliability, and generalizability tests in experiments, internal consistency in any study is a strong indicator of descriptive psychometric integrity but contributes to the examination of test-prone items (APA, 2019). Another matching study is using transformers carried out by Kaur and Arora in Int. J. Scientific Research in CS (vol. 10. no. 5. 2024. pp. 112- 118). [14]R. Gupta, S. Bhatt, and M. Rao, “Enhancing Recruitment Processes using Hybrid NLP Models,” Journal of Intelligent Systems and Applications, vol. 18, no. 1, pp. 22- 28, 2024. A hybrid NLP article by Gupta, Bhatt, and Rao is published in Journal of Intelligent Systems and Applications (vol18 no1 pp 22-28). [15]A. Singh, Resume Entity Recognition with Custom SpaCy Pipelines, Journal of machine learning and NLP vol. 5, no. 3, pp. 39-45, 2023. Singh refers to custom SpA Cy as discussed in the Journal of Machine Learning and NLP (vol 5, no. 3, 2023, pp 39) -45. [16]Y. Li, L. Liu, and J. Huang, “NER to parse a resume with the use of transfer learning techniques,” Computational Linguistics Research, vol. 7, no. 2, pp. 56-63, 2023. The transfer learning NER can be found in the paper of Li, Liu, and Huang in the publication Computational Linguistics Research (vol 7, no 2, 2023, pp 56-63). [17]P. Reddy and S. Mishra, “Comparative Analysis of Resume Parsing Tools on the Structured Data Mining and Calculation on a – Based data, IJERA, vol. 14, no. 1, pp. 88 92. The analysers of parsing tools are assessed by Reddy and Mishra (vol. 14, no. 1, 2023, pp. 88-92). [18]M. Shah and D. Banerjee, From Keywords to Concepts: Semantic Matching in Recruitment International Journal of Computer Applications Technology and Research, vol. 12 no. 6 pp. 115-120, 2025. The article by Shah and Banerjee on the semantic matching in the year 2025 is published in the Int. J. Computer Applications Technology and Research (vol 12, no 6, pp 115-120). [19]B. Kapoor and R. Sethi, “Explainable AI in Hiring: SHAP Values of Model Interpretability,” IEEE Transactions on Human Machine Systems, vol. 52, no. 4, pp. 300 -309, 2024. In an IEEE Human-Machine Systems paper (vol 52, no. 4 pp. 300-309), Kapoor and Sethi address SHAP-based explanations. [20]Staffing Industry Analysts, “India White -Collar Recruitment Surges 19% in September, Market Report, 2025. According to the report by the Staffing Industry, the hiring in the white-collar sector in India increased 19 percentinSeptember,2025.

© 2025 International Journal of Engineering and Techniques (IJET).

Digital Object Identifier (DOI)

IJET assigns a unique DOI to every accepted article via Zenodo for persistent linking and citation. Your article’s DOI: https://doi.org/{{doi}}

Open DOI About Zenodo DOI

Close