We introduce Oscar, a personalized educational assistant serving as a companion to young students to maximize the quality of education and enable them to learn at their own pace. Our formulation uses a comprehensive set of attributes that are needed to be modelled to provide individualized learning to students. We discover these attributes using an ensemble of three LLMs. We then use the attribute-based profile to build a GenAI solution for individualized learning. Through manual human annotation, we identify that only 19\% common attributes are provided by all three LLMs, while 31.5\% are common across two LLMs; and the remaining 49.5\% are specific to only one of them; demonstrating diverging understanding across LLMs. Utilizing a consolidated attribute profile, Oscar displays highly customized responses to individual student needs. We discuss the strengths and limitations of the approach and offer recommendations for educators, GenAI developers as well as policymakers to promote the integration of GenAI tools in childhood education.