NYIT developed an LLM assistant for an applied analysis course in MATLAB

Researchers from the Department of Electrical Engineering and Computer Technology at the New York Institute of Technology (NYIT) presented in July 2026 on arXiv VectorizationLLM — a narrowly specialized language assistant based on open-weight models from Google, developed for one specific educational course.

For which course was the assistant created

VectorizationLLM was developed exclusively for course CTEC 247 — "Applied Computational Analysis II" — of the ECE Technology Department at NYIT's Old Westbury, New York campus. The course covers several topics that are traditionally difficult for students in technical specialties:

• Smart vectorization in the MATLAB environment
• Analysis of time and wave vectors
• Piecewise-defined functions
• Fourier analysis
• Ordinary differential equations

In these topics, students most often get stuck and seek help outside the classroom. The assistant fills this gap, being available around the clock — unlike an instructor in the department.

How the model works — and what it doesn't do

The architecture of VectorizationLLM is built on two components. The first is RAG (Retrieval Augmented Generation): a knowledge base formed from the materials of course CTEC 247, including lecture notes, MATLAB code examples and illustrations. When a student asks a question, the model first searches for relevant fragments within this corpus, then generates an answer based on them. This dramatically reduces the risk of "hallucinations" compared to universal chatbots.

The second component is a system prompt that defines the pedagogical role: the assistant explains and hints, but intentionally does not provide ready-made solutions to tasks. The principle of "Socratic dialogue" — the student should reach the answer independently. Responses can contain text explanations, code blocks and images simultaneously, which allows explaining mathematical concepts in the most visual format.

Why "course AI" is a promising direction

VectorizationLLM fits into a growing trend of creating narrowly specialized educational LLMs tied to a specific subject or course. Unlike ChatGPT in "raw" form, such an assistant does not go beyond the curriculum, relies only on verified instructor materials and reduces the burden on the department in terms of routine student questions.

The key challenge when scaling similar solutions is populating the RAG corpus: creating a quality knowledge base for each course requires time and academic expertise. This, rather than the LLM itself, is where the main work is concentrated.

What this means

VectorizationLLM shows how Google's open-weight models can be applied in education without expensive fine-tuning: the RAG + system prompt combination allows quickly adapting the base LLM to any course while maintaining control over content. Widespread adoption of such solutions in universities will depend on how easily instructors can independently form such knowledge bases.