Work

Fitscout: A Retrieval-Augmented Generation Based AI Platform for Intelligent Candidate Screening and Ranking

Year: 2026
Type: preprint
Abstract: Hiring the right candidate has always been a complex task, and the tools most organizations rely on today — keyword-based Applicant Tracking Systems — make it harder than it needs to be. These systems filter resumes by matching words, not meaning, which means genuinely qualified people often get screened out while others who have simply learned to game the system slip through. This paper introduces Fitscout, a recruitment platform that takes a fundamentally different approach. Built around Retrieval-Augmented Generation (RAG), semantic vector search, and Large Language Models, Fitscout evaluates candidates the way a thoughtful recruiter would — by understanding context, not just counting keywords. The system pulls together structured data like GPA and experience alongside unstructured resume text, then uses that combined picture to generate ranked candidate lists, plain-language hiring verdicts, and cohort-level insights. Under the hood, it runs on FastAPI, ChromaDB, Sentence Transformers, and the Groq-hosted Kimi K2 model for fast inference. When tested against traditional ML baselines, the RAG pipeline showed a clear advantage in contextual reasoning and role-specific ranking accuracy.
Source: HAL
Language: English
Open Access status: green
OpenAlex ID: W7234269371
Cites:
Cited by:
Related to: 10