This paper introduces Askademia, an Artificial Intelligence (AI) system designed to provide real-time, context-aware responses to electronically submitted student questions during live lectures. Traditional approaches often fail to deliver timely answers, especially in large courses. Askademia employs a novel Vision-Language Model (VLM) to respond to student questions based on live lecture content and retrieved course materials. We evaluated Askademia on student questions from a large-scale university course over one semester. We compared the quality of Askademia’s responses with those provided by Teaching Assistants (TAs). Using both quantitative metrics and expert human evaluations, we assessed responses for relevance, readability, factuality, and tone. The results demonstrate that Askademia delivers high-quality answers with significantly greater speed and consistency. Our analysis showed that 98.3% of Askademia’s responses arrived before a topic shift, with a median response time of \(2\,\textrm{s}\) , compared to 45.4% of TA responses, which had a median of \(241\,\textrm{s}\) . Moreover, our human evaluation metrics showed that Askademia’s responses were high quality, with strong readability, relevance, and factual accuracy scores.