In the era of rapid digital transformation, continuous professional development and skill enhancement have become imperative. This paper presents a novel software prototype designed to support certification exam preparation through personalized learning pathways powered by advanced Machine Learning techniques. The system comprises a comprehensive multi-platform ecosystem (web, mobile, and desktop applications) enabling users to complete competency assessments via adaptive multiple-choice questionnaires.The primary contribution of this work lies in the development of an automated question classification system coupled with an intelligent adaptive test generation module, implemented within a scalable and resilient software architecture. The AI-powered classifier, developed in Python, employs a multi-provider aggregation strategy that synthesizes outputs from multiple Large Language Model services—OpenAI GPT, Anthropic Claude, Google Gemini, and Cohere—augmented by dedicated APIs and conventional Natural Language Processing techniques (TF-IDF, spaCy). This hybrid approach ensures robust item analysis and precise categorization aligned with educational learning objectives. The system incorporates sophisticated fallback mechanisms, Redis-based caching, and comprehensive cost monitoring capabilities. The software architecture features a Java-based backend implementing RESTful services interfaced with a MySQL relational database. Asynchronous task processing is handled through Celery workers, while the monitoring infrastructure leverages Prometheus, Grafana, and the ELK Stack for comprehensive observability. The entire solution is containerized using Docker and supports orchestration via Kubernetes, ensuring horizontal scalability and high availability. To optimize the learning experience, an adaptive test generation algorithm dynamically constructs assessments comprising 70-80 questions while respecting multiple constraints including difficulty progression, topic distribution, and examination duration. The algorithm continuously monitors individual learner performance, iteratively adjusting content delivery by detecting learning plateaus, identifying knowledge gaps, and dynamically calibrating question difficulty to maximize pedagogical effectiveness and accelerate competency acquisition.