Loading content…
QBS Co.
In-house agentic HR candidate screening platform at QBS Co. automating bulk resume evaluation, semantic candidate-job matching, and LLM fine-tuning via knowledge distillation.
PythonRAGC#.NET CoreLangChainLLMsKnowledge DistillationVector SearchAgentic Workflows
ResumeHR is an in-house agentic AI platform engineered for QBS Co. to automate bulk resume screening and match candidates to open job roles with high precision. Handling high volumes of candidate submissions, the platform parses unstructured CVs and resumes, extracts structured professional profiles (competencies, career progression, skills, educational credentials), and semantically ranks applicants against specific job descriptions.
To achieve fast, high-throughput inference without the latency and high API costs of large frontier models, the platform utilizes Knowledge Distillation. Reasoning and ranking capabilities from larger teacher models are distilled into compact, fine-tuned student LLMs specifically optimized for resume-job description alignment. The system features a hybrid Retrieval-Augmented Generation (RAG) pipeline combined with LangChain agentic workflows, supported by a robust C# / .NET Core backend for bulk file ingestion, document storage, and recruiter dashboards.
Manual screening of hundreds of bulk applicant resumes is time-consuming, prone to human fatigue, and costly when relying on standard commercial LLM API calls per document.
Architected an agentic pipeline using Python, LangChain, and RAG for dense vector semantic matching, coupled with knowledge distillation to fine-tune a compact, low-latency student model for candidate scoring. Built the enterprise backend and document ingestion services using C# and .NET Core.
Automated end-to-end bulk resume processing at QBS Co., delivering instant candidate-job semantic ranking, explainable relevance breakdowns for HR teams, and dramatic reductions in inference latency and cost via distilled models.
Skills
