Pengembangan Virtual HRD Berbasis LLM Untuk Latihan Wawancara Kerja Yang Interaktif

Nirwansyah, Reza Ali (2025) Pengembangan Virtual HRD Berbasis LLM Untuk Latihan Wawancara Kerja Yang Interaktif. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024211001-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy

Abstract

Wawancara kerja merupakan tahap penting dalam proses rekrutmen, namun banyak fresh graduate merasa kurang siap menghadapinya. Penelitian ini mengembangkan sistem simulasi wawancara kerja berbasis Large Language Model (LLM) dengan arsitektur multi-agent. Sistem terdiri dari empat agent utama: Question Agent untuk membuat pertanyaan, Asker Agent untuk menyampaikan pertanyaan, Summary Agent untuk merangkum jawaban, dan Review Agent untuk mengevaluasi kelengkapan wawancara. Implementasi menggunakan React.js untuk frontend, Flask untuk backend, dan MongoDB untuk penyimpanan data. Pengujian dilakukan dengan membandingkan model OpenAI GPT-4.1 mini dan Gemini 2.5 Pro pada skenario wawancara posisi Fullstack Developer. Hasil menunjukkan sistem mampu menghasilkan wawancara yang natural dan kontekstual, dengan Gemini 2.5 Pro lebih unggul dalam pendekatan holistik dibandingkan GPT-4.1 mini yang lebih fokus teknis.
======================================================================================================================================
Job interviews are an important stage in the recruitment process, yet many fresh graduates feel unprepared to face them. This research develops a job interview simulation system based on Large Language Model (LLM) with multi-agent architecture. The system consists of four main agents: Question Agent for creating questions, Asker Agent for delivering questions, Summary Agent for summarizing answers, and Review Agent for evaluating interview completeness. Implementation uses React.js for frontend, Flask for backend, and MongoDB for data storage. Testing was conducted by comparing OpenAI GPT-4.1 mini and Gemini 2.5 Pro models in Fullstack Developer position interview scenarios. Results show the system can generate natural and contextual interviews, with Gemini 2.5 Pro being superior in holistic approach compared to GPT-4.1 mini which is more technically focused.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Simulasi Wawancara, multi-agent, Large Language Model, Virtual Interviewer, AI recruitment Interview Simulation, multi-agent, Large Language Model, Virtual Interviewer, AI recruitment
Subjects:	Q Science > QA Mathematics > QA76.758 Software engineering
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Reza Ali Nirwansyah
Date Deposited:	29 Jul 2025 05:39
Last Modified:	29 Jul 2025 05:39
URI:	http://repository.its.ac.id/id/eprint/122269

Actions (login required)

View Item