The rapid digitalization of higher education demands efficient tools to evaluate multi-dimensional academic performance. Traditional evaluation methods are limited by high time consumption, siloed data structures, and an over-reliance on static numeric aggregates.This paper presents the architectural design, algorithmic implementation, and empirical validation of an AI-driven academic analytics module integrated into Kherson State University's Virtual Learning Environment, "KSU24". The module automates the aggregation, sanitization, and qualitative synthesis of heterogeneous faculty performance data across 14 activity domains.Integrated with leading LLMs (GPT-4o, Gemini 1.5 Pro, Claude 3.5 Sonnet) via a deterministic software pipeline, the platform transforms relational database entities into actionable narrative profiles within a semi-automated Human-in-the-loop governance structure. To prevent hallucinations, we implement role-restricted prompting, XML-bounded context windows, and schema-driven data ingestion.Field testing over 346 authentic faculty records confirmed 0.00% factual error rates and reduced report composition time from 40–60 minutes down to 2–5 minutes per profile. Finally, we formalize the semantic translation layer and propose a career trajectory simulation framework for strategic institutional forecasting.