SkillSync: An Explainable AI Framework for Resume Evaluation, Skill Gap Analysis, and Career Alignment

SkillSync: An Explainable AI Framework for Resume Evaluation, Skill Gap Analysis, and Career Alignment

Jeffson Preetham Dsilva Dept. of Computer Applications,

St. Joseph Engineering College, Mangaluru, India

Murari B K,

Dept. of Computer Applications,

St. Joseph Engineering College, Mangaluru, India

Abstract – Resume screening has evolved in the modern recruitment process, with the integration of Applicant Tracking Systems (ATS). But job seekers continue to face challenges in aligning their resumes with job-specific requirements. This paper presents SkillSync, an AI-powered resume analysis platform that gives intelligent feedback to improve the employability chances of the candidate. The system evaluates resumes using both rule-based and semantic models, offering ATS-style scoring, formatting suggestions, skill gap analysis, and AI-generated cover letters. It includes sentence-transformer-based embeddings to semantically match resume content with the skills required for the selected job role derived from the datasets such as O*NET and LLaMA2- formatted job roles. In addition, SkillSync leverages Gemini AI to give real-time explanations for missing skills and to assist users with an chatbot for resume and job-related queries. The system recommends personalized upskilling resources from Coursera, NPTEL, and YouTube using API-based retrieval and duration- based filtering. This paper demonstrates that SkillSync bridges the gap between static resumes and dynamic industry needs through a comprehensive, explainable, and scalable framework. The proposed system aims to standardize access to career-readiness tools, particularly benefiting students and early-career professionals.

Index TermsResume Evaluation, Skill Gap Analysis, Semantic Similarity, AI-based Career Guidance

1. INTRODUCTION

   IN today's competitive job market, an effective resume is essential for the chances of candidates selected. Recruiters often receive hundreds of resumes for one job hiring. This makes it difficult to evaluate each candidate fairly and efficiently. Thus, organizations have started using Automated Tracking Systems (ATS) and smart resume screening tools to improve candidate selection [1]. But still, many of these systems rely on keyword matching and strict templates, that too frequently fail to capture the context of a candidate's experience or the applicability of skills to evolving industry demands.

   As more companies utilize Natural Language Processing (NLP) and Artificial Intelligence (AI) in Human Resource Management, improved resume analysis software has emerged. They measure formatting quality, skill appropriateness, and job

   matching through AI-based methods like semantic similarity, named entity recognition, and transformer-based models [2][3]. Even with these advancements, there is still a vast gap in tools that can improve the employability of a candidate. These

   tools should offer:

   • An explainable skill gap analysis,

   • Suggestions for learning content tailored to a candidate's missing skills,

   • AI-generated resume summaries and cover letters,

   • Interactive guidance using conversational AI.

   To cover this gap, I am introducing SkillSync, an AI- powered resume analyzer that evaluates a candidate's resume based on formatting, relevance, and skills. It uses semantic matching with Sentence-BERT, provides an explainable ATS scoring system, and uses Gemini AI for cover letter creation and skills feedback. Additionally, it connects to real-time data from platforms like Coursera, NPTEL, and YouTube API to suggest relevant upskilling resources. SkillSync also has a chatbot interface provided by Gemini, offering career guidance and resume advice in a conversational way.

   This project aims to create a clear, adaptable, and smart career navigation tool for students and professionals. It fills the gap between candidates' skills and what industries expect in today's AI-driven recruitment landscape.

2. RELATED WORK

   The use of Artificial Inteligence (AI) in evaluating resumes and matching job roles is a growing field of research. Many important studies have looked into the difficulties of automating candidate screening and improving Applicant Tracking Systems (ATS).

   Luo et al. introduced ResumeNet, a learning-based model designed to assess resume quality using hierarchical neural networks. This system highlights the need for structured resume representation to improve recruiter decisions [1]. Similarly, Li et al. [2] proposed a transformer-based architecture that predicts candidate competence levels and matches resumes to job descriptions using contextual embeddings. Their work insists

   the importance of semantic understanding over traditional keyword matching.

   Bevara et al. developed Resume2Vec, a framework that employs intelligent resume embeddings to enhance candidate matching in ATS. Their model significantly improves precision in identifying suitable candidates [3]. Other works like those by Tayal et al. [4] and Jafari [5] demonstrate the application of machine learning clasifiers, like SVM and Random Forest, for classification of resumes based on job-fit scores.

   Saatci et al. [6] explored the application of Natural Language Processing (NLP) for resume screening, focusing on extracting relevant keywords and assessing resume structure. Sun et al. [7] proposed context aware sentence similarity approaches using BERT, enabling systems to recognize paraphrased skill expressions. Their findings were then supported by Reimers and Gurevychs Sentence-BERT, which optimized semantic similarity through Siamese and triplet network structures [13]. In addressing ethical concerns, Hofeditz et al. [8] introduced Explainable AI (XAI) in recruitment systems, aiming to prevent bias and improve transparency in screening tools.

   While most of these systems focus on one or two aspects like ATS scoring, formatting checks, or job matching, this proposed system SkillSync integrates multiple modules, including resume parsing, ATS scoring with semantic similarity, formatting evaluations, course recommendation, and LLM- based AI explanations.

3. METHODOLOGY

   The proposed system, SkillSync, is a AI-driven platform designed to assess and improve resumes based on the candidates desired job role. The system follows a layered methodology combining resume parsing, semantic skill matching, ATS-style scoring, and personalized recommendations. The workflow includes resume text extraction, formatting assessment, keyword matching, semantic similarity analysis, and AI-generated insights. To support different user requirements, the system includes external APIs for Google Gemini, YouTube, and includes domain-specific datasets (LLaMA2 fine-tuned roles and O*NET occupational knowledge). The entire platform is implemented as an Streamlit web application, optimized for usability and modular analysis.

   1. System Overview

      The proposed system, SkillSync, is designed as a comprehensive AI-driven platform that assist job seeker in optimizing their resumes for desired job roles. It follows a modular architecture composed of several integrated components, where each address a specific functionality within the resume evaluation and guidance workflow.

      The system begins with the Resume Parsing Module, which extracts plain text from uploaded PDF resumes. This raw text is later processed by the Skill Extraction Module, which uses regular expressions and part-of-speech heuristics to identify candidate skills. These are matched semantically against a curated set of required skills for the specificroles, leveraging pre-trained sentence embedding models like MiniLM-L6 and cosine similarity scoring. This enables contextual matching, which outperforms traditional keyword matching by understanding the intent and semantics of skill expressions [3].

      The ATS Scoring Module evaluate resume based on real- world recruiter preferences. This includes parameters like keyword presence, relevant job title, tool mentions, and section coverage. Simultaneously, the Formatting Evaluation Engine checks for structural issues such as a lack of section headers, use of images/tables and inconsistent formatting. To support the upskilling journey of candidate, the Recommendation Module provides relavant courses from Coursera, NPTEL, and YouTube, related to missing skills identified in the resume. Also, SkillSync integrates Google Gemini AI to offer two advanced features:

      • Natural Language Explanations for missing skills: Why this skill matters in this role

      • Personalized AI-generated Cover Letters, depending on the resume content and the target job.

        All modules are integrated within a Streamlet web interface, allowing real-time feedback, visualization via interactive charts, and a floating AI chatbot for conversational guidance. The complete pipeline is shown in the system architecture diagram (Fig. 1).

        Fig. 1. System architecture of skillsync

   2. Dataset Preparation

      The proposed system depends on two main data sources:

      1. the O*NET occupational database and

      2. a custom LLaMA2-formatted dataset developed for job- role-to-skill alignment.

      The O*NET (Occupational Information Network) dataset is a public database maintained by the government which offers information about job roles, required skills, tasks, and knowledge areas for different professions. The Excel files found on the official O*NET Resource Centre website [18] are used in this system. I have parsed these files using Python's pandas and openpyxl libraries and converted them to JSON format with a custom script (merge_onet_descriptions.py). The extracted fields from these files include Job Title, Skills, Knowledge, Tasks, Description, and Occupation. This process helped seamless integration with the system's job-role analyser while maintaining data from the original source. I have also created a custom dataset formatted for LLaMA2-style prompt- based models, which contains job roles, required skills, projects, and a summary of jobs in plain language. This dataset was manually compiled and structured to show realistic expectations in both IT and non-IT roles.

      I have merged both datasets after removing duplicate roles and aligning skill terminology. Roles found only in O*NET were added to the final dataset to improve coverage across different sectors. This combined dataset serves as the knowledge base for skill gap analysis, resume matching, and

      personalized recommendations throughout the system.

   3. Functional Workflow

      The SkillSync system follows a modular and well-defined pipeline that ensure seamless flow of data from resume upload to feedback generation using Gemini AI. The functional architecture is shown in Fig. 2. The core workflow components are:

      1. Resume Upload: Users interact with the SkillSync interface using a web-based Streamlit application. Upon launching, the user uploads a resume in .pdf format. The system temporarily stores the file and prepares it for parsing.

      2. Text Parsing: The uploaded PDF resume is parsed using the PyMuPDF library, which extracts plain text. This textual content serves as input to the other modules. This phase also involves noise filtering and whitespace normalisation to make sure clean input for downstream analysis.

      3. Semantic Skill Extraction: Instead of keyword matching, SkillSync uses semantic embedding techniques. Resume content and job-specific required skills are encoded using the all-MiniLM-L6-v2 Sentence Transformer [13]. The cosine similarity between the resume vector and each required skill vector is computed. A skill is marked as matched if its similarity score crosses a defined threshold (0.55), and otherwise classified as missing. This provides more accurate recognition of contextually implied skills [13], even if keywords differ.

      4. Scoring Engine: The resume is evaluated on multiple things: ATS Score: This evaluates keyword match, formatting, section coverage, metrics, and tool usage based on

         industrial standards [1][3].

         Skill Match: Calculated depending on the ratio of matched to required skills using semantic similarity.

         Each score is visualised using gauges and pie charts using Plotly for better interpretability.

      5. Gemini Feedback (LLM Integration)

      To improve explainability and personalisation, the Gemini Pro LLM is integrated through Googles Generative AI API. The output is contextual, descriptive, and human-like, offering value-added feedback [12][20].

      Fig. 2. Sequence Diagram of Resume Processing Pipeline

   4. User Workflow & Scenarios

      SkillSync is designed with a primary focus on students, actively preparing for internships or job applications. The use

      case diagram (Fig. 3.) outlines the interactions between the user and the systems core functionalities.

      1. Actor: User/Student; The user is the sole actor in this system. Their primary interactions with this platform include these.

      2. Upload Resume: Upload a .pdf resume file for evaluation.

      3. Select Job Role: Choose a desired job profile from a predefined list

      4. Get Evaluation Feedback: Get feedback on ATS score, formatting score, skill match and gap analysis.

      5. Generate Cover Letter: Automatically generate a customized cover letter using Gemini AI, based on the resume content and selected role.

      6. View Learning Recommendations: Discover curated learning resources from Coursera, NPTEL, and YouTube based on skill gaps.

      7. Chat with Gemini AI Assistant: Interact through a built-in chatbot for career related queries and guidance.

      These interactions are illustrated in Fig. 3.

      Fig. 3. Use Case Diagram for SkillSync

      User Interaction Flow:

      The flow of user interaction in SkillSync is designed to be streamlined for students and early-career professionals to analyse their resumes efficiently. The flow begins when a user uploads their resume and selects a job role. The system then processes the resume through various analysis modules, displays scores, suggests learning resources, and allows AI- based interactions like generating cover letter and chatbot conversations.

      The main flow of user interaction is described below:

      1. Start

      2. Upload Resume (.pdf)

      3. Select Desired Job Role

      4. Resume Text is Extracted

      5. Formatting and ATS Scoring Performed

      6. Skill Extraction and Semantic Matching

      7. Feedback Displayed: Scores, Charts, Skill Gap

      8. Recommendations Generated (Courses, Videos)

      9. User Generates Cover Letter (Gemini AI)

      10. User May Chat with Gemini Chatbot

      11. End / Repeat

      This interaction process is visually represented in Fig. 4.

      evaluations provide a ATS score that reflects relevance, quality, and domain fit.

      1. Skill Gap Detection Module: SkillSync intelligently identifies skill gaps by comparing resume-derived skills against the expected skill sets for a given job role. The system extracts skill entities from resume text using a combination of regular expression heuristics and noun phrase detection. It then applies semantic matching through sentence-transformers to identify matching skills and missing skills, based on a similarity threshold. This module quantifies the users skill alignment, showing the percentae of skills that meet the job requirements and visualizes the gaps for targeted upskilling.

      2. Recommendation Engine: Based on the detected skill gaps, this module recommends relevant learning resources. It integrates with the YouTube Data API, filtering long-form tutorials for quality content. Additionally, it auto-generates links to Coursera and NPTEL course search results using direct keyword mapping. Users receive domain-specific content for each missing skill. This module not only supports self-learning but also addresses industry-demanded competencies in a personalized way.

      3. Gemini AI Integration: SkillSync uses Google's Gemini Pro

      2.5 API for natural language generation tasks. Gemini is used to:

      Fig. 4. Activity Diagram showing User Interaction

   5. Modules Description

   SkillSync is composed of several coordinated modules that operate in sequence to analyze, evaluate, and enhance a candidates resume.

   1. Resume Parsing Module: This module processes resumes uploaded in PDF format and extracts clean text using PyMuPDF. The extracted text becomes the major input for all other modules.

   2. Formatting Checker: To assess the readability and professional structure of a resume, this module analyses various formatting features. It checks for the presence of crucial sections such as "Experience", "Projects", "Education", and "Skills". Also, it flags problematic formatting elements like unwanted use of capital letters, missing section headers, and overall text balance. The output is a Formatting Score (out of 100) along with precise suggestions, enabling users to align with industry-standard formatting conventions. The implementation is done using Pythons re module for pattern detection and rule-based logic.

   3. ATS Scoring Engine: This module simulates the behavior of modern Applicant Tracking Systems (ATS) that recruiters use to filter resumes. It performs scoring using both rule-based and semantic methods. First, it uses keyword matching to score the presence of predefined technical and soft skills. Then, it provides semantic similarity scoring using embeddings generated by Sentence-BERT (MiniLM model) and calculates cosine similarity between resume content and role-specific descriptions or skill sets [1][3][13]. These two layered

   Summarize uploaded resumes in coherent narratives. Explain the relevance and application of missing skills. Generate role-specific, customized cover letters.

   The integration provides explainable AI feedback, helping users to understand and improve their resume and skillset with clarity [8][12]. The use of LLMs for targeted content generation makes the platform adaptive and future-ready.

   7) Interactive Chatbot: An embedded chatbot powered by Gemini AI provides a conversational interface. It allows users to interact with the system in real time, asking career advice, resume queries, or feedback explanations. The chatbot is styled as a floating icon on the interface and utilizes Geminis conversational capabilities to deliver accurate and personalised responses, providing user engagement and interactivity.

4. RESULTS

1. Evaluation Metrics Used

   To check the quality and relevance of resumes, the system uses a hybrid scoring methodology that combines keyword- based scoring with semantic similarity analysis. The following metrics were used:

   1. ATS Score (out of 100): Depending on keyword matching, section coverage, tools and certifications, metrics used, job role relevance, and grammar.

   2. Formatting Score (out of 100): Depending on formatting guidelines like section presence, font uniformity, absence of all-caps, etc.

   3. Skill Match (%): Percentage of required job role skills found in the resume using semantic cosine similarity ([13]).

   4. Visual Score Representation: Scorecards and gauges generated using Plotly provide quick feedback to users.

2. Output Snapshots

   Below are visual outputs generated by SkillSync:

   1. ATS & Formatting Scores: Gauge charts display both ATS and formatting scores, making it easier for users to interpret results at a glance.

      Fig. 5. Visualization of ATS and Formatting Scores

   2. Skill Match Pie Chart: A donut chart shows the percentage of matched vs. missing skills for the selected job role.

      Fig. 6. Visualization of Skills Matching Percentage

      Metric	Value
      ATS Score	89 / 100
      Formatting Score	85 / 100
      Skill Match	55%
      Missing Skills	Laravel, CodeIgniter, REST API, Symfony
      Matched Skills	PHP, MySQL, JavaScript, HTML, CSS
      AI Suggestions Provided	Yes
      Cover Letter Generated	Yes

   3. Gemini AI Skill Explanations: Missing skills are explained using Gemini AI. These explanations are personalized to the chosen job role and help users understand their skill gaps.

      Example: Skill: REST API

      Fig. 7. Gemini AI Explanation for Missing Skills

   4. Resume Summary: The system generates an AI- written resume summary that reflects the applicants profile using generative AI.

      Fig. 8. AI Generated Resume Summary

   5. Cover Letter Generation: Based on the resume content and selected role, the system generates a professional cover letter using Gemini AI.

   Fig. 9. Cover Letter Generated by Gemini AI

3. Case Study: Resume Evaluation For Php Developer

   Fig. 10. Output for PHP Developer Role

   This result shows that the resume is well-structured and has strong relevance to the selected role but still lacks some critical backend or deployment-related skills.

4. Course Recommendations

   To bridge the skill gap identified during resume analysis, SkillSync integrates learning resource recommendations using external APIs:

   1. YouTube API: Fetches long-form tutorials (minimum 30 minutes) for each missing skill.

   2. Coursera and NPTEL: Direct course links are generated based on the missing skill keywords.

   Fig. 12. Course Recommendations for missing skills

5. Gemini Chatbot

   To improve user engagement and provide personalized guidance, SkillSync adds Googles Gemini AI as a conversational assistant within the user interface. The chatbot offers real-time, context-aware support for a variety of user questions related to careers and resumes. Key use cases like resume improvement queries, career advice, skill explanations etc.

   E.g., What is Laravel and why is it important for a PHP developer?

   Fig. 13. Gemini AI Chatbot

   V. DISCUSSIONS

   The implementation of SkillSync shows how AI and Natural Language Processing can meaningfully help job seekers in improving their resumes and aligning them with industry expectations. This section analyses the broader implications, strengths, limitations, and potential future improvements of the system.

   1. Effectiveness of Hybrid Scoring: The combination of keyword-based scoring and semantic similarity (cosine- based) provides a more precise evaluation of a resumes relevance to a selected job role. While keyword scoring ensures alignment with ATS that depend on exact matches, semantic scoring helps capture contextual relevance. This method increases the adaptability across domains and reduces false negatives in skill matching, as demonstrted in prior literature [1][2].

   2. Enhancing Transparency and User Trust: Adding Gemini AI serves two purposes:

      Explainable AI (XAI): It enables users to understand why certain skills are missing or important, aligning with calls for transparency in intelligent hiring systems [8].

      Human-Like Interaction: A conversational interface reduces the entry barrier for users with less technical background, making the system accessible.

   3. Strengths:

      Modular Design: Each module is separately testable and maintainable.

      Cross-Domain Capability: By adding two datasets (LLaMA2-based and O*NET), the system supports both technical and non-technical roles.

      Up-to-date Skill Recommendations: Real- time integration with platforms like Coursera, NPTEL, and YouTube ensures users access modern, relevant learning paths.

   4. Limitations

      Lack of Resume Formatting PDF Parsing: Though the text is extracted, the loss of layout may affect format-specific evaluation like headers or columns.

      Gemini API Quota Dependency: Overuse or rate-limiting of the Gemini API can affect continuous availability.

      Limited Non-English Support: The system currently performs best with English resumes and may not be useful for other languages.

   5. Ethical and Fairness Considerations: Systems that score resumes must be evaluated for bias and fairness.

SkillSync mitigates this by:

Avoiding demographic-based filters

Using semantic matching to prevent penalizing users for alternate phrasing

Providing explanations for missing skills, rather than just penalizing them

However, future work should explore fairness metrics and bias audits more formally [8].

1. CONCLUSION

   In this study, I have introduced SkillSync, an AI-powered resume evaluation and recommendation system designed to bridge the gap between candidate profiles and evolving industry requirements. By leveraging both keyword-based ATS scoring and semantic similarity techniques, SkillSync delivers a more accurate and intelligent assessment of resumes. The system contains key modules such as resume parsing, formatting analysis, semantic skill gap detection, and course recommendations from Coursera, NPTEL, and YouTube. Also, the integration of Google Gemini AI enables advanced functionalities including automated resume summarizing, missing skill explanation, and AI-generated cover letters. An interactive chatbot built into the interface supports continuous user engagement and guidance throughout the application.

   The modular and scalable design of SkillSync makes it suitable for deployment in educational institutions, placement cells, and professional development platforms. As the next step, the system can be extended to support many resume formats, introduce multilingual support for broader accessibility, integrate with platforms like LinkedIn for real-time profile analysis, and add bias detection mechanisms to ensure fairness in resume evaluation. With ongoing refinement, SkillSync has the potential to evolve into a comprehensive ecosystem for personalized, intelligent, and unbiased career guidance and enhancement.

2. REFERENCES

1. Luo, Yong & Zhang, Huaizheng & Wang, Yongjie & We, Yonggang & Zhang, Xinwen. (2018). ResumeNet: A Learning-based Framework for Automatic Resume Quality Assessment. 10.48550/arXiv.1810.02832.

2. Li, Changmao & Fisher, Elaine & Thomas, Rebecca & Pittard, Stephen & Hertzberg, Vicki & Choi, Jinho. (2020). Competence-Level Prediction and Resume & Job Description Matching Using Context-Aware Transformer Models.

3. Bevara, R.V.K.; Mannuru, N.R.; Karedla, S.P.; Lund, B.; Xiao, T.; Pasem, H.; Dronavalli, S.C.; Rupeshkumar, S. Resume2Vec: Transforming Applicant Tracking Systems with Intelligent Resume Embeddings for Precise Candidate Matching. Electronics 2025, 14, 794. https://doi.org/10.3390/electronics14040794

4. Dr. Sandeep Tayal, T. Sharma, S. Singhal, and A. Thakur, Resume Screening using Machine Learning, Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol, vol. 10, no. 2, pp. 602606, Apr. 2024, doi: 10.32628/CSEIT2410275.

5. Jafari, Faizan. (2025). Automated Resume Screening System Using NLP and Machine Learning. INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT. 09. 1-9. 10.55041/IJSREM48082.

6. Saatci, M., Kaya, R., & Ünlü, R. (2024). Resume Screening With Natural Language Processing (NLP). Alphanumeric Journal, 12(2), 121-140. https://doi.org/10.17093/alphanumeric.1536577.

7. Xiaofei Sun, Yuxian Meng, Xiang Ao, Fei Wu, Tianwei Zhang, Jiwei Li, Chun Fan; Sentence Similarity Based on Contexts. Transactions of

   the Association for Computational Linguistics 2022, 10 573588. doi: https://doi.org/10.1162/tacl_a_00477.

8. Hofeditz, L., Clausen, S., Rieß, A. et al. Applying XAI to an AI-based system for candidate management to mitigate bias and discrimination in hiring. Electron Markets 32, 22072233 (2022). https://doi.org/10.1007/s12525-022-00600-9

9. International Journal of Scientific Research in Computer Science, E., & IJSRCSEIT, I. T. (2022). Resume Screening using Machine Learning and NLP: A proposed system. International Journal of Scientific Research in Computer Science, Engineering and Information Technology. https://doi.org/10.32628/CSEIT228240

10. Aakankshu Rawat, Siddharth Malik, Seema Rawat et al. A Systematic Literature Review (SLR) On The Beginning of Resume Parsing in HR Recruitment Process & SMART Advancements in Chronological Order, 01 July 2021, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-570370/v1]

11. Chavan, Prasad & Chandurkar, Yash & Tidake, Ankita & Lavankar, Gaurav & Gaikwad, Suhani & Chavan, Rohit. (2024). Enhancing recruitment efficiency: An advanced Applicant Tracking System (ATS). Industrial Management Advances. 2. 6373. 10.59429/ima.v2i1.6373.

12. Dhruv Awasthi, Pooja S. B. Rao, and Dinesh Babu Jayagopi. 2025. ResumeGenAI: Supporting Job Seekers with LLM-Driven Resume Feedback. In Proceedings of the 7th ACM Conference on Conversational User Interfaces (CUI '25). Association for Computing Machinery, New York, NY, USA, Article 21, 19. https://doi.org/10.1145/3719160.3737622

13. Reimers, Nils & Gurevych, Iryna. (2019). Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. 10.48550/arXiv.1908.10084.

14. Bhatia, V., Rawat, P., Kumar, A., & Shah, R. R. (2019). End-to-end resume parsing and finding candidates for a job description using bert. arXiv preprint arXiv:1910.03089.

15. Heakl, A., Mohamed, Y., Mohamed, N., Elsharkawy, A., & Zaky, A. (2024). ResuméAtlas: Revisiting Resume Classification with Large- Scale Datasets and Large Language Models. Procedia Computer Science, 244, 158-165.

16. Gan, C., Zhang, Q., & Mori, T. (2024). Application of llm agents in recruitment: A novel framework for resume screening. arXiv preprint arXiv:2401.08315.

17. Resume Optimisation and Suggestions using Large Language Models: A Review. Authors: Sai Ganesh Mandala, Ruthvik Reddy Korem, Karthik Karmakonda. https://doi.org/10.22214/ijraset.2024.59069

18. National Center for O*NET Development. O*NET Database. Accessed:

    June 2025. https://www.onetcenter.org/database.html

19. Jeffson Dsilva. Custom LLaMA2-style job-skill dataset for SkillSync. GitHub Repository. Accessed: June 2025. https://github.com/Jeffson- dsilva/SkillSync

20. Ohm, Akshansh & Krishna, Bhavani. (2019). Chatbot for Career Guidance Using AI. International Journal of Computer Sciences and Engineering. 7. 856-860. 10.26438/ijcse/v7i6.856860.

21. D. P. B, M. K, V. K and Y. D, "Job and Coure Recommendation System using Collaborative Filtering and Naive Bayes algorithms," 2023 2nd International Conference on Advancements in Electrical, Electronics, Communication, Computing and Automation (ICAECA), Coimbatore, India, 2023, pp. 1-4, doi: 10.1109/ICAECA56562.2023.10200758