InternHub: A Resume-Driven Internship & Job Recommendation Framework with Integrated Scam Screening, Application Tracking and Carrer & Skill Gap Analytics

Abstract

Students seeking internships face bloated repositories of postings, frequently plagued by cross-platform duplicates, expired vacancies, and fraudulent stipend schemes. Contemporary literature on internship recommendation primarily concentrates on ranking algorithms under the assumption that pre-structured candidate profiles already exist and that all listings are verified. This paper presents InternHub, an end-to-end web framework that addresses the entire internship discovery and management pipeline. A student uploads an unstructured PDF resume; a dual-condition negative heuristic filter intercepts non-resume documents (e.g., college tuition fee receipts and financial bills); and a natural language processing pipeline extracts a canonical taxonomy of technical skills, project experiences, and target role classifications. In parallel, internship postings are retrieved concurrently across five live job APIs (JobDataAPI, Jooble, Adzuna, Remotive Open API, and Arbeitnow), transformed to a unified schema, pruned of dead links via mirror synchronization, and de-duplicated using composite tuple keys. A rule-based Trust Score Engine screens postings for scam indicators, penalizing upfront fees, unverified chat channels, and truncated specifications. Surviving opportunities are scored and ranked via a mathematically bounded five-factor model (core skills 50%, role alignment 20%, preferred skills 15%, project implementation evidence 10%, work-mode/location preferences 5%), retaining TF-IDF cosine similarity as a lexical baseline. For each opportunity, InternHub conducts a skill-gap analysis, suggests targeted resume refinements, and generates a progressive four-phase learning roadmap. Application states are followed across a six-stage lifecycle (Saved, Applied, Interview, Selected, Rejected, Withdrawn). The platform is implemented using React 19, TypeScript, Express.js, a FastAPI NLP microservice (PyMuPDF, spaCy, scikit-learn), and Supabase PostgreSQL with Row-Level Security (RLS). Functional verification across an 18-scenario automated suite achieved a 100% pass rate, with full end-to-end recommendation requests averaging 1.84 s.

Introduction

The paper introduces InternHub, an intelligent internship recommendation platform designed to simplify the internship search process for engineering students. Traditional job portals require students to manually enter their details, filter through thousands of irrelevant opportunities, verify job authenticity, and identify missing skills. InternHub addresses these challenges by combining Artificial Intelligence (AI), Natural Language Processing (NLP), machine learning, and automated security checks into one platform.

1. Existing Problems

Traditional internship recommendation systems have several limitations:

• Students must manually enter their academic and technical details.

• Job portals contain irrelevant, duplicate, expired, and fraudulent listings.

• Existing systems mainly focus on matching candidates with jobs rather than identifying skill gaps.

• Students must independently track applications and prepare for required competencies.

2. Proposed System – InternHub

InternHub provides an integrated solution through five major features:

1. Automated Resume Parsing: Extracts education, technical skills, projects, and experience from PDF, DOCX, and TXT resumes using NLP. It also rejects irrelevant documents such as invoices and fee receipts.

2. Scam Detection and Trust Scoring: Collects internships from five external job providers, removes duplicates, and assigns each listing a Trust Score between 15 and 98 to identify potentially fraudulent opportunities.

3. Explainable Job Recommendations: Uses five factors—technical skills, job-title alignment, preferred qualifications, project evidence, and candidate preferences—to calculate personalized match scores.

4. Skill Gap Analysis: Identifies missing competencies required for selected internship roles and generates a four-phase learning roadmap covering Foundations, Core Frameworks, Advanced Tools, and Deployment.

5. Application Tracking: Allows students to save opportunities and monitor application progress through six states.

3. System Architecture

InternHub uses a decoupled microservice architecture consisting of:

• Frontend: React 19, TypeScript, Vite, and Tailwind CSS.

• Backend: Node.js and Express for authentication, API integration, aggregation, and recommendation processing.

• NLP Service: Python, FastAPI, PyMuPDF, spaCy, and scikit-learn for resume parsing and text analysis.

• Database: Supabase PostgreSQL for storing profiles, applications, and saved internships.

• Security: JWT authentication and Row-Level Security (RLS) to prevent unauthorized access to student information.

4. Methodology

The system follows these major steps:

1. Students upload their resumes.

2. The NLP engine extracts and standardizes technical skills.

3. Internship listings are collected concurrently from five job providers.

4. Duplicate, expired, and suspicious listings are filtered.

5. A Trust Score is calculated to identify potentially fraudulent postings.

6. A five-factor recommendation algorithm ranks internships according to candidate suitability.

7. Missing skills are identified, and a personalized learning roadmap is generated.

8. Students save opportunities and track their application progress.

TF-IDF and cosine similarity are also implemented as baseline recommendation techniques for comparison.

5. Experimental Results

The system was evaluated using component tests, performance measurements, and automated functional testing.

These results demonstrate that the platform's implemented components function as intended. However, the reported scam detection performance is based on a limited, artificially constructed dataset rather than real-world fraudulent listings.

6. Limitations

Despite its capabilities, InternHub has several limitations:

• Skill extraction depends on a predefined technical skill dictionary.

• Trust Score rules require validation using larger real-world datasets.

• Exact-key matching may fail to identify differently worded duplicate listings.

• Recommendation accuracy has not yet been validated against independent human judgments.

• External API availability and response times can affect performance.

• Resume formatting differences may introduce bias.

• Student information requires strong privacy protection and data deletion options.

Conclusion

InternHub combines resume ingestion, non-resume rejection, multi-provider aggregation, scam screening, explainable five-factor scoring, skill-gap guidance and application tracking in one platform. All 18 functional cases passed, and a full recommendation request averaged 1.84 s with live APIs. The Trust Score and filter results come from test sets built by the authors, so the next step is the evaluation in Section V-E.

References

[1] F. Ricci, L. Rokach, and B. Shapira, Eds., Recommender Systems Handbook, 2nd ed. New York, NY, USA: Springer, 2015. [2] G. Adomavicius and A. Tuzhilin, “Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions,” IEEE Trans. Knowl. Data Eng., vol. 17, no. 6, pp. 734–749, Jun. 2005. [3] B. S. Geethanjali, S. Karol, P. M. Reddy, R. V. P. Ganesh, and R. J. Bharathi, “AI-based internship recommendation system,” in Proc. 10th Int. Conf. Smart Struct. Syst. (ICSSS), 2025, doi: 10.1109/ICSSS66939.2025.11346221. [4] “TechTrack: An internship matching platform for career recommendation using machine learning,” in Proc. 11th Int. Conf. Bus. Ind. Res. (ICBIR), 2026, doi: 10.1109/ICBIR69932.2026.11584730. [5] S. Mol S. and M. Manoj, “AI-driven internship matching using skill-based recommendation system,” in Proc. Nat. Conf. Emerg. Comput. Appl. (NCECA), vol. 7, no. 1, Amal Jyothi College of Engineering, 2026. [6] D. R. S. Victoria, L. Gokulraj, R. Hem Kumar, S. Roselin Mary, and S. Nithya Roseline, “AI-based internship recommendation engine for the PM Internship Scheme,” in Proc. Int. Conf. Emerg. Smart Inf. Commun. Technol. (ICESIC), 2026, doi: 10.1109/ICESIC67389.2026.11496549. [7] A. Busad, G. U. Shankar, S. Dhanush, D. Abbigeri, N. Patil, and K. M. Darshana, “Smart internship recommender,” Int. Res. J. Mod. Eng. Technol. Sci. (IRJMETS), vol. 8, no. 1, pp. 2733–2739, Jan. 2026. [8] S. Vidros, C. Kolias, G. Kambourakis, and L. Akoglu, “Automatic detection of online recruitment frauds: Characteristics, methods, and a public dataset,” Future Internet, vol. 9, no. 1, art. 6, 2017. [9] B. Alghamdi and F. Alharby, “An intelligent model for online recruitment fraud detection,” J. Inf. Secur., vol. 10, no. 3, pp. 155–176, 2019. [10] M. J. Pazzani and D. Billsus, “Content-based recommendation systems,” in The Adaptive Web, P. Brusilovsky, A. Kobsa, and W. Nejdl, Eds. Berlin, Germany: Springer, 2007, pp. 325–341. [11] P. Lops, M. de Gemmis, and G. Semeraro, “Content-based recommender systems: State of the art and trends,” in Recommender Systems Handbook. Boston, MA, USA: Springer, 2011, pp. 73–105. [12] G. Salton and C. Buckley, “Term-weighting approaches in automatic text retrieval,” Inf. Process. Manage., vol. 24, no. 5, pp. 513–523, 1988. [13] K. Spärck Jones, “A statistical interpretation of term specificity and its application in retrieval,” J. Documentation, vol. 28, no. 1, pp. 11–21, 1972. [14] C. D. Manning, P. Raghavan, and H. Schütze, Introduction to Information Retrieval. Cambridge, U.K.: Cambridge Univ. Press, 2008. [15] A. I. Schein, A. Popescul, L. H. Ungar, and D. M. Pennock, “Methods and metrics for cold-start recommendations,” in Proc. 25th ACM SIGIR Conf., 2002, pp. 253–260. [16] B. Sarwar, G. Karypis, J. Konstan, and J. Riedl, “Item-based collaborative filtering recommendation algorithms,” in Proc. 10th Int. Conf. World Wide Web (WWW), 2001, pp. 285–295. [17] R. Burke, “Hybrid recommender systems: Survey and experiments,” User Model. User-Adapt. Interact., vol. 12, no. 4, pp. 331–370, 2002. [18] M. Jones, J. Bradley, and N. Sakimura, “JSON Web Token (JWT),” IETF, RFC 7519, May 2015. [19] Supabase Documentation. [Online]. Available: https://supabase.com/docs [20] PostgreSQL Documentation. [Online]. Available: https://www.postgresql.org/docs [21] Adzuna API Documentation. [Online]. Available: https://developer.adzuna.com [22] Jooble REST API Documentation. [Online]. Available: https://jooble.org/api/about [23] Remotive API Documentation. [Online]. Available: https://remotive.com/api [24] JobDataAPI Documentation. [Online]. Available: https://jobdataapi.com [25] Arbeitnow Job Board API Documentation. [Online]. Available: https://www.arbeitnow.com/api/job-board-api [26] FastAPI Documentation. [Online]. Available: https://fastapi.tiangolo.com [27] spaCy Documentation. [Online]. Available: https://spacy.io [28] PyMuPDF Documentation. [Online]. Available: https://pymupdf.readthedocs.io [29] F. Pedregosa et al., “Scikit-learn: Machine learning in Python,” J. Mach. Learn. Res., vol. 12, pp. 2825–2830, 2011.

Copyright

Copyright © 2026 Mr. Aaksath Nithil Saam SS, Mrs. Jayabalajivel R, Mrs. Yoheswari S. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.