TalentLoop – AI-Driven Resume Generation and Semantic Verification via Live Codebase Integration and Dynamic ATS Weighting
Computer and Decision Making An International Journal · 04.11.2026
DOI: 10.1922/comdem.v3i6s.19151 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making An International Journal, online erschienen am 04.11.2026. Verfasst von Aryan Naik und 5 weiteren Autor:innen.
Abstract (Originalsprache)
In modern hiring practices, the standard Applicant Tracking System (ATS) is dependent on manipulative keyword-based algorithms, making them susceptible to the creation of false credentials and technical skills. In this paper, we present TalentLoop, a novel Tri-Score recruitment architecture that abandons static heuristics in favor of dynamic machine learning and live semantic verification. To eliminate the subjective black-box scoring, a Random Forest Regressor was trained on a proprietary dataset of 1,128 mathematically pristine, API-verified developer resumes. This model dynamically weighs 14 distinct structural features, such as action-verb density and Type-Token Ratio; to evaluate formatting integrity. To verify the truthfulness of technical claims, TalentLoop utilizes a local, Parameter-Efficient Fine-Tuned (PEFT) Llama 3.1 8B agent to asynchronously cross-reference resume content against live GitHub API codebase data. Additionally, an API-grounded Retrieval-Augmented Generation (RAG) builder strictly enforces JSON schemas to generate verified, high-scoring resumes. Adversarial benchmarking demonstrates that while leading commercial ATS platforms blindly validate injected 404 project links, TalentLoop’s architecture successfully identifies and penalizes fraudulent claims. Moreover, the fine-tuned agentic engine for generation significantly raised the median performance score of genuine candidates from an unoptimized score of 86.0 to a score of 93.5. The structural evaluation happens almost in real-time (101.49±188.34ms) while agentic verification runs in the background effectively (74.60±43.84s). This study provides a robust, data-driven mechanism for verifiable automated recruitment.
Zitieren
APA
Naik, A., Ranjan, A., Jadhav, D., Chhabria, A., Gulwani, R., & Deshpande, R. (2026). TalentLoop – AI-Driven Resume Generation and Semantic Verification via Live Codebase Integration and Dynamic ATS Weighting. Computer and Decision Making An International Journal. https://doi.org/10.1922/comdem.v3i6s.19151
Vancouver
Naik A, Ranjan A, Jadhav D, Chhabria A, Gulwani R, Deshpande R. TalentLoop – AI-Driven Resume Generation and Semantic Verification via Live Codebase Integration and Dynamic ATS Weighting. Computer and Decision Making An International Journal. 2026. doi:10.1922/comdem.v3i6s.19151
BibTeX
@article{Naik2026,
author = {Naik, Aryan and Ranjan, Apeksha and Jadhav, Dipti and Chhabria, Aditi and Gulwani, Reshma and Deshpande, Renuka},
title = {TalentLoop – AI-Driven Resume Generation and Semantic Verification via Live Codebase Integration and Dynamic ATS Weighting},
journal = {Computer and Decision Making An International Journal},
year = {2026},
doi = {10.1922/comdem.v3i6s.19151}
}
EUVIMED